March, 2026 - Faz Business | فاز الأعمال

Best SAP AI Integration Services For Smart Automation

Posted on March 20, 2026 by faz_business

SAP AI Integration Services: Connecting Your SAP Environment to Enterprise AI

Where Most SAP AI Projects Actually Break?

An enterprise spends three months selecting an AI vendor, six weeks scoping the use case, and then hits a wall: the AI system and the SAP environment are not talking to each other the way anyone expected. Data pipelines stall. API authentication fails in the production environment. The model produces outputs that make no sense because it is reading the wrong SAP table.

SAP AI integration is where most enterprise AI programs lose momentum. Not in the model selection. Not in the use case design. In the connection layer between the AI capability and the SAP data and workflows it needs to be useful.

USM Business Systems is a specialized SAP AI delivery partner headquartered in Ashburn, VA. We integrate enterprise AI systems — LLMs, agentic frameworks, predictive models — into live SAP environments for manufacturers, pharma companies, logistics operators, and the system integrators that serve them.

What SAP AI Integration Actually Covers?

SAP AI integration is not a single service. It spans five distinct layers, and the difficulty of each depends on your SAP landscape, your data maturity, and the AI capability you are connecting.

Data Layer Integration

Before any AI system can reason accurately about your SAP environment, it needs a clean, structured feed of the right data. This typically means connecting to SAP Datasphere (SAP’s data fabric), SAP HANA views, or extracting structured data from S/4HANA tables using OData APIs or SAP Data Services.

The most common failure point here is master data quality. AI models amplify whatever is in your data. If your material master has inconsistent UoM coding across plants, a demand forecasting model will surface that inconsistency as erratic predictions.

API and Middleware Integration

Most enterprise AI integration with SAP runs through SAP BTP Integration Suite — SAP’s managed integration platform that handles API management, protocol translation, and event streaming between SAP and external systems. Engineers who have not worked with BTP Integration Suite before underestimate the configuration depth it requires, particularly for high-volume transactional workflows.

AI Runtime Integration

SAP AI Core is the managed runtime where enterprise AI models are deployed, versioned, and governed inside the SAP ecosystem. Integrating an external LLM or a custom predictive model into SAP AI Core requires specific API patterns, credential management, and lifecycle configuration that differs from deploying the same model in AWS or Azure. SAP AI Core engineers — not general ML engineers — are the right resource here.

Workflow and Process Integration

An AI capability that produces a recommendation but cannot act on it is a dashboard, not an integration. Real SAP AI integration connects the AI output back into SAP workflows: a quality prediction that triggers a production hold in SAP PP, a demand signal that adjusts a replenishment order in SAP IBP, a document analysis result that routes an invoice exception in SAP Finance.

User Experience Integration

For AI capabilities that surface to end users inside SAP, integration with SAP Fiori and SAP Joule determines whether the capability gets adopted. Engineers who understand both the AI layer and the SAP UX layer are required. These are not the same people.

What is the fastest path to a production SAP AI integration?

The fastest path starts with a single, well-scoped workflow that has clean source data in SAP. A supplier performance monitoring integration or an invoice exception routing integration can reach production in 8-12 weeks when the data is ready. Broad integrations that touch multiple SAP modules simultaneously take 4-6 months minimum.

Can we integrate a third-party LLM — like GPT-4 or Claude — directly into SAP?

Yes. SAP AI Core supports external model connections, and SAP BTP Integration Suite handles the API management layer. The integration work involves authentication, data formatting, latency management, and governance configuration. This is a well-established integration pattern for document analysis, NLP search, and content generation use cases.

The Three Integration Patterns We See Most Often

Pattern 1: NLP Search on SAP Data

Enterprises add a natural language search layer on top of SAP Datasphere or HANA, allowing users to query supply chain, financial, or operational data in plain language rather than through SAP transaction codes. According to Forrester’s 2024 Enterprise AI Survey, 61% of SAP users report that data accessibility is the primary barrier to AI adoption. NLP search directly addresses this.

The integration connects an LLM to SAP data views, with a retrieval layer that fetches relevant records and passes them to the model as context. The model returns an answer in plain language. The SAP Fiori interface surfaces the result. This pattern reaches production in 6-10 weeks for a defined data domain.

Pattern 2: Document AI on SAP-Connected Document Flows

Enterprises processing high volumes of documents — invoices, purchase orders, quality certificates, compliance filings — integrate document AI to extract, classify, and route content automatically. The integration reads documents from SAP Document Management or external repositories, processes them through a document AI model, and writes the structured output back to the relevant SAP object.

Pharma and life sciences companies use this pattern for batch record processing and supplier qualification documents. Logistics companies use it for freight invoice reconciliation. The accuracy rate on standard document types typically reaches 90%+ within the first 30 days of production operation.

Pattern 3: Predictive Models on SAP Operational Data

Predictive models trained on historical SAP transaction data — demand history, equipment sensor readings, supplier delivery records — produce forward-looking signals that feed back into SAP planning processes. A demand forecasting model reads S/4HANA sales history and external market signals, produces a forecast, and updates SAP IBP automatically. A predictive maintenance model reads equipment telemetry and writes a maintenance recommendation to SAP PM.

This pattern has the longest data preparation phase — 4-8 weeks to clean and structure SAP historical data — but produces the highest sustained value once in production.

What to Look for When Evaluating SAP AI Integration Partners

SAP AI Core and BTP Integration Suite experience, specifically. Ask for examples of integrations built on these platforms, not SAP integrations in general.
Data readiness assessment as part of the scoping process. Partners who jump straight to architecture without assessing your SAP master data quality are skipping the step that determines whether the integration will work.
A clear governance model. Enterprise SAP environments are audited. Any AI integration needs logging, version control, human override capability, and a rollback procedure.
Engineers who have worked in both the AI layer and the SAP layer. The rarest and most valuable profile is an engineer who understands SAP data structures and modern AI frameworks simultaneously. Firms that staff these roles separately add significant coordination overhead.

Why USM Business Systems?

USM Business Systems is a CMMi Level 3, Oracle Gold Partner AI and IT services firm headquartered in Ashburn, VA. With 1,000+ engineers, 2,000+ delivered applications, and 27 years of enterprise delivery experience, USM specializes in AI implementation for supply chain, pharma, manufacturing, and SAP environments. Our SAP AI practice places specialized engineers inside enterprise programs within days — on contract, as dedicated delivery pods, or on a project basis.

Ready to put SAP AI into production? Book a 30-minute scoping call with our SAP AI team at usmsystems.com.

Get In Touch!

FAQ

How does SAP BTP Integration Suite differ from standard API middleware?

BTP Integration Suite is SAP’s managed platform for enterprise integration — it handles API management, event streaming, protocol translation, and pre-built connectors to SAP and third-party systems. It also integrates directly with SAP AI Core, which is what makes it the preferred integration layer for SAP AI programs.

What data from SAP can be used to train AI models?

Historical transactional data from S/4HANA, master data from SAP MDG, sensor data connected through SAP IoT, and document data from SAP Document Management are all commonly used. The key requirement is data governance — understanding what data can leave SAP boundaries and what must stay in the SAP environment.

How long does a SAP AI integration project take from scoping to production?

A single, well-defined integration — one workflow, one AI capability, one SAP module — typically takes 8-14 weeks from scoping to production deployment. Multi-module integrations or programs that require significant data preparation first run 4-6 months.

What is SAP Datasphere and why does it matter for AI integration?

SAP Datasphere is SAP’s data fabric platform — it creates a unified, governed data layer across SAP and non-SAP sources. For AI integration, it is important because it gives AI models a clean, semantically structured view of enterprise data without requiring direct access to S/4HANA tables.

Can AI integrations be built incrementally, or do they require a full platform build first?

Incremental is the right approach for most enterprises. A first integration scoped to one workflow proves the pattern, builds internal confidence, and reveals integration requirements you did not anticipate. Enterprises that try to build a complete AI integration platform before demonstrating value rarely reach production.

The Friday Roundup – Gear Envy and Speed Ramping Tips

Posted on March 20, 2026 by faz_business

Pop art cartoon image of man buy too much camera gear.

Amateur Camera Gear Not Worth The Money in 2026

Although the title of the video below refers to “Camera Gear,” it actually goes way broader than that and is well worth watching.

The harsh reality is that in the past four or five years the changes or improvements to most cameras has been incremental and mainly focused on tiny details.

The obsession with upgrading yearly is almost entirely a false concept that is created and driven by marketing departments all over the world.

Let’s face it, most tech companies whether they are hardware or software based are built on a model that depends on new products being released annually.

That can be a new model camera or an updated version of some software but either way their profits depend on you and I buying into the latest and greatest.

How to Make Speed Ramps Actually Work

Most reasonably equipped video editing software these days will offer some kind of dedicated Speed Ramping feature.

If you are not sure what speed ramping is, it is the action of speeding up or slowing down portions of a clip to create dynamic interest.

Anyway, given that the tools necessary are widely available most people use speed ramping in a pretty basic way.

That’s fine but as is usually the case, that’s not exactly what the professionals are doing!

If you check out the video below you will see that using or not using speed ramping is actually not the whole story.

The real point is how, why and exactly where you apply speed ramping if at all to your edits.

Use PowerDirector as a Photo Editor

I came across this video demonstration this week and was surprised that a lot of people were not aware of this.

Then I realized that using a video editor as an image editor was a relatively new concept given that the control available in older video editing software wasn’t all that good.

However in the past two years or so the amount of color adjustment and control you can exercise in your average video editor has improved out of sight.

So here’s a look at what you can do with a still image in the current version of PowerDirector.

Beginner’s Guide to AI Image to Video in PowerDirector (2026)

So right now one thing that is happening in the world of A.I. as applied to video is that we are seeing more and more separate A.I. models emerging.

Whilst I guess that is a good thing from a competition and pricing perspective it also adds a degree of confusion to the subject.

The problem is that each model will have strengths and weaknesses in performing certain tasks.

In this video from CyberLink you can see what I am talking about as they go through an example of “image to video” A.I. generation for different models.

Edit Your Podcast Yourself (No Studio Needed)

An interesting “crossover” that has been going on for the past few years is the idea of podcasts being distributed as videos.

In fact some of the bigger podcasts at this point in time now have three for four camera setups and pro lighting on top of the high end mic setups!

The reasoning behind this is quite simple, has proven to be successful over time and follows a very old rule.

That rule is: “Don’t just be on the internet, be everywhere on the internet!”

So in the case of podcasts the idea is you upload that podcast to every possible service you can and then, you upload the video to every possible service as well.

Of course editing your own podcast/video at home can be quite a good deal of work if you don’t have a process.

So, here’s a process you can follow to keep things simple and efficient.

30-Minute Video Editing Challenge

Over the past few years there has been a heavy emphasis in the world of video editing software on the development of A.I.

Unfortunately when things like this enter just about any space the respective marketing departments take over and all of a sudden we are in the realm of wizzbang, ooh-ahh look at this cool thing!

The real value of A.I. in video editing is not its ability to create videos but its ability to add useful tools to make your job as a video editor easier.

A.I. cannot really “create” anything, it will always need a human to tell it what to create!

Here’s a video from the people at Filmora showing how A.I. tools can be used to help you as a creator work more efficiently.

10 Gimbal Moves to Get Cinematic Footage

If you have gotten yourself a gimbal in search of that buttery smooth handheld footage there’s something that you are probably now aware of!

First up is the fact that just loading that camera or phone onto the gimbal is not going to guarantee smooth motion especially when walking.

Turns out you still have to adjust the way you move to give the gimbal a chance.

The second point is that the gimbal is absolutely not going to capture any decent footage!

In a strange twist the real story is that you have to shoot those cinematic masterpieces and the gimbal is actually only a tool! Who knew!

Anyway, here are some tips to help you lift that gimbal game.

Ask me Anything about YouTube

So here’s the deal. If you are in any way serious or want to get serious about running a YouTube channel professionally (not just a hobby!) then you need to watch this video.

Having said that let me provide a little context.

First up you will need to set aside about two hours to watch it!

It is a Q&A so there are a lot of topics covered but more importantly it is from Derral Eves who is probably one of the most successful YouTube consultants around at the moment.

In fact Derral has been around for quite a number of years now and is an authority that is well worth listening to.

Everyone Overcomplicates Color Grading

One thing I have noticed over the years of putting together the Friday Roundup is that in some ways the constant addition of information on the internet is actually counter intuitive.

What I mean by that is that as more and more information is added a an apparancy of complexity begins to rear its ugly head!

One such subject is that of color grading although the phenomena is not restricted to that alone!

In the video below Casey walks through this subject with the specific goal of simplifying the entire process and making it way more approachable.

Discover more from The DIY Video Editor

Subscribe to get the latest posts sent to your email.

Android Canary update brings big changes, but nothing is guaranteed

Posted on March 20, 2026 by faz_business

Google has rolled out a fresh Android Canary build, and it looks more complete than you might expect from an experimental release. The catch is simple. This version exists to test ideas, not to promise them.

Android Canary 2603 bundles practical additions like app lock, chat bubbles, and a redesigned screen recorder. It’s available across a wide range of Pixel devices, but this update is meant for developers rather than everyday use

That context matters when reading into any of these changes. Features shown here can still be removed before a stable release, even if some eventually appear in beta builds.

Even so, this build offers a useful snapshot of where Android may be heading next.

New features worth watching

App lock is one of the most practical additions. You can now long press an app and secure it directly, which adds a layer of privacy without digging through settings menus

That same menu also activates bubbles more fully. Conversations can float on screen as overlays, making multitasking feel more natural than before

The screen recorder has also been reworked into a floating pill interface. It lets you quickly choose between recording the full screen or a single app, then moves you into a preview flow where you can edit, delete, or share the clip

Google has even refined smaller details. The long press menu now groups shortcuts into a cleaner layout that expands only when needed.

Why this update isn’t final

There’s a reason none of this should be taken as final. This channel exists for developers to explore early features and APIs, not as a preview of what will ship next

That gap between polish and certainty is important. Some features already feel complete, but they can still be adjusted or dropped before Android 17 reaches stability. Even changes that move into beta are not guaranteed to stick.

There’s also a tradeoff to consider. Canary builds aren’t designed for daily use, so most people won’t try them unless they’re comfortable dealing with bugs or manual installs.

What you’re seeing here is Android in an active state of change, with ideas being tested in public before decisions are locked.

What to expect next

Some of the smaller changes hint at a broader direction. The return of separate Wi-Fi and mobile data toggles suggests a rethink of earlier design decisions, while heavier use of blur points to ongoing visual refinement

There’s still no clear timeline for what carries forward. Google hasn’t confirmed which of these features will land in Android 17 or when they might reach stable devices, even on supported Pixel models.

If you’re tracking Android’s evolution, the next step is to watch what shows up in beta builds. That’s usually where experimental ideas start turning into features you’ll actually use.

Best 50 Sites to Hire Freelance CAD Design Experts and Remote Designers for Your Company

Posted on March 20, 2026 by faz_business

In the highly competitive landscape of the design industries, businesses from all around the world are in tough competition not only in terms of sales but also in finding skilled CAD design experts. Things are even more difficult for startups and small businesses, as they might not be able to afford an in-house professional design team from the get-go, so they tend to rely on an external workforce to complete specific design tasks. It’s not that they don’t have a full-time employee to do the job – it’s just that hiring a freelancer or two can help get things done without all the complexities of permanent or contractual recruitment.

And that’s just the tip of the iceberg of advantages. Hiring 3D CAD design freelancers means you have on-demand access to experts/professionals with specialized design skills that your team is currently lacking. Since the freelancer isn’t part of your payroll, you’re under no obligation to keep the person hired after the project is done. Most freelancers offer their services on a per-project basis. It’s cost-effective because you avoid expenses like lengthy training, onboarding, office space, employee benefits, etc.

The temporary nature of freelancing makes it easier for you to scale the business workforce up and down depending on the current necessities. With the ability to quickly adapt to the constantly changing needs comes the benefit of reduced hiring risk. There’s no higher turnover rate, affording you the stability within the in-house team and a work environment conducive to higher productivity. Listed below are some of the best online resources to help you find and hire CAD drafting professionals and remote designers.

1. Cad Crowd

Most freelancing websites have a pretty broad scope of categories, ranging from web development and marketing to administrative support and accounting services. Cad Crowd is unlike the vast majority of freelancing sites. In fact, it’s one of the very few that specializes in CAD (computer-aided design), MEP (Mechanical engineering services, Electrical, Plumbing), and AEC (Architectural, Engineering, and Construction). If your business has anything to do with design and engineering, and you need a skilled freelancer to get something done, Cad Crowd is likely the only place you should go to.

Powered by more than 125,000 professionals from various backgrounds in the design and engineering industries, you’ll be hard-pressed to NOT find the right person for any design job. You have several options to discover talents in Cad Crowd. The most straightforward route is to get a quote directly from the site; to do that, it’s important to provide some details about your project, including length and type. Add some images, which can be samples or project files, if necessary. Based on your description, Cad Crowd will provide a quote from a pre-qualified professional – this is part of their “Managed Services” feature.

Alternatively, the site allows you to post a “Design Contest” to elicit submissions from dozens of CAD designers and CAD engineering professionals within a matter of days. A design contest might be public for everyone to see and participate, private (accessible by only select designers), or invite-only in case you want submissions from invited users. Another highlight is that Cad Crowd manages everything from the early submission process to NDA agreement and payment methods. Even if you need to hire freelancers on an hourly basis, it’s also possible via their managed services. In general, Cad Crowd is a one-stop shop for your freelance hiring needs in the CAD design industry.

Website: Cadcrowd.com

2. X-PRO CAD

There doesn’t seem to be any option for you to browse and directly hire freelancers on X-PRO CAD. But you can provide short details about a project and ask for an estimate. The site specializes in everything CAD, including but not limited to engineering design, animation, and consumer product design. It offers patent filing and related services if you want those as well. Interestingly enough, X-PRO CAD doubles as a prototype maker and manufacturing partner; the idea is that once you have the design ready in a CAD file, they can take on the production process using 3D printing, CNC machining, or injection molding methods.

Website: X-procad.com

3. Indeed

Nowhere on the site does Indeed think of itself as a freelancing website. It’s a place where job seekers can browse through employment opportunities posted by organizations and companies. Indeed seems to avoid the use of the term “freelancers” in favor of “temporary” and “contract,” in addition to the usual full-time and part-time options. Therefore, if you insist on using Indeed to look for and hire freelancers, make sure your job posting is categorized under the contract or temporary listing. Remember that a freelance CAD job can be regarded as any of the two types, anyway. There really isn’t any glaring difference.

Website: Indeed.com

4. Worksome

Again, you can’t hire directly from the site, but Worksome offers a platform to broadcast a job opening, whether freelance in nature or otherwise, through a “multichannel” distribution network. While it all sounds sophisticated, this basically refers to publishing the project or job on many different platforms. This should help broaden the audience reach and increase the exposure in general. As usual, you’ll be able to customize the posting itself, such as by defining the specific roles, skill sets, and locations, among others. Worksome is also a platform for freelance management that gives you a complete overview of the hired workforce, including payments and contracts.

Website: Worksome.com

5. Glassdoor

Using Glassdoor to attract CAD drafting freelancers for any given project is quite unusual, but not a far-fetched idea either. The site is built mostly for job seekers as they browse through employers’ profiles and decide which companies to apply for. A neat trick to get the attention of the right talent is to craft your business description in a way that tells them how you often find yourself in need of freelancers to keep up with the workload. Considering the fact that the vast majority of Glassdoor users are active job seekers, it certainly is worth a shot. However, this might only work if your business occasionally needs freelancers. For a one-time project, Glassdoor is less than ideal.

Website: Glassdoor.com

6. Unicorn Factory

If you’re willing to hire remote workers from anywhere in the world for your next CAD design project, then Unicorn Factory can definitely cater to your needs. But if your company can only hire remote workers from specific countries or territories, bear in mind that the site can only connect you with freelancers from two countries: New Zealand and Canada. You have three options to find freelancers here: post a job, browse the directory, contact the listed freelancers directly, or use the concierge service. In any case, Unicorn Factory claims that most clients should find the right freelancers within just a few days.

Website: Unicornfactor.nz

7. Contra

Unlike a lot of freelancing platforms that often boast about their ability to cater to businesses of all sizes, Contra uniquely positions itself as a tool for businesses currently struggling to find the much-needed talent. It’s not a job board or marketplace for freelancers, but a platform for companies to source contract workers (or freelancers), manage project workflow, and process payments. The online application has just about everything you need to make hiring easier and managing projects more practical. Contra actually invites the most experienced freelancers to join their team for priority hiring. You can also use some pre-made contract templates suitable for different freelancer hiring scenarios as well.

Website: Contra.com

8. Aquent Talent

Who says you can’t use a recruitment agency to hire freelancers? It’s certainly a possibility, and there’s nothing wrong with that at all. One of those agencies is Aquent Talent. Although recruitment agencies are often associated with permanent, or at least long-term staffing, Aquent Talent offers recruitment services for quick projects, which translates to hiring contractual workers, in other words, freelancers. Think of it as hiring vetted 3D design freelancers through a staffing agency that will manage everything in the process.

Website: Aquenttalent.com

9. Nexxt

Like the Aquent Talent mentioned above, Nexxt is also a recruitment agency. One of the biggest differences between the two is that Nexxt actually allows you to post a job (for a fee, of course) and have the platform manage it for better targeted exposure. The job posting is published across Nexxt’s own network of career and local freelancing sites, so the project quickly gains a sizable online presence, all in the right places. There is also the “Programmatic Advertising” plan, where you have more features like job posting management, ROI measurement, budget optimization, and exposure increase.

Website: Nexxt.com

10. Kolabtree

Here is the site to find freelancers with respectable credentials. Kolabtree is basically similar to most freelancing sites, except for one thing. You can easily post a job or a project and expect to see some proposals from freelancers within the next few days. The distinguishing aspect is that Kolabtree claims to only allow freelancers with impressive backgrounds, such as an Ivy League education, to apply for any project. In case you want to contact some freelancers directly without posting a project, Kolabtree makes it easy for you to browse their profiles as well. All payments are processed through an escrow system.

Website: Kolabtree.com

11. DesignCrowd

As long as your next CAD project revolves around graphic design services, make sure to give DesignCrowd a go. While every project is packaged in a “design contest” environment, it doesn’t change the fact that the site is obviously all about freelancing. It starts with you initiating a design contest, and then interested freelancers begin to submit their designs according to the project brief. You pick the best design and reward the winner with prize money. DesignCrowd is a nice place to launch a “test project” and see whether the site actually has skillful freelancers to work on your next, larger job. If a contest is too time-consuming, you might want to contact the freelancers directly instead.

Website: Designcrowd.com

12. FlexJobs

An employer’s account in FlexJobs is not free. The most basic account starts at $199, which already comes with unlimited job postings and unlimited resume searches. According to FlexJobs itself, about 82% of the job seekers registered with the site have a college degree, and nearly 75% have worked remote jobs in some capacity, either part-time or full-time. Since it was established in 2007, FlexJobs has helped connect 4 million job seekers in their search for flexible employment options. It doesn’t say if the platform specializes in any field or industry, but it only means the platform accepts job postings for any project, including CAD design.

Website: Flexjobs.com

13. Remote.co

Established in 2015, Remote.co is a sister site of FlexJobs. Both platforms offer more or less the same thing, but Remote.co focuses on remote hiring. In case your business has no problem with hiring CAD professionals from everywhere in the world (some companies do have restrictions about this),Remote.co is definitely worth taking a look at. There seem to be quite a lot of CAD freelancers in various sectors, such as 3D animators, architectural drafters, concept artists, product designers, etc.

Website: Remote.co

14. Guru

A general category freelancing site, Guru is a place where hundreds of businesses post their projects for free. You can also boost the listing for better visibility and greater exposure for a relatively small fee. Freelance (and remote) CAD designers can respond to the job posting by providing bids and giving you quotes for their services. You have the option to set your own contract with the selected freelancer, whether with an hourly rate or a fixed price. A contract for recurring work is also available. Once the project is underway, you might use Guru’s own project management tool called “WorkRooms” for collaboration and communication purposes. Payment is done with an escrow method for safety.

Website: Guru.com

15. 99designs

Primarily a graphic design contest platform, 99designs offers a simple, straightforward, quick way to discover a freelancer for your next project. To hire a designer, browse through their listing of freelancers and invite the one you like to collaborate on a project. Every progress, deliverable, and communication is done within the site. Once the design is finished, you release the payment, and the project concludes. Or, you can choose the traditional route of posting a project and launching a public contest. The latter might be preferable, but only for a simple project, where you don’t have to go back and forth with every designer for revisions and additional details.

Website: 99designs.com

17. Behance

Owned by Adobe, Behance is first and foremost a marketplace for CAD assets. Accordingly, you need an Adobe account to access all features on the site. Before you start searching for freelancers in Behance, it might be a good idea to determine the timeline and budget for the project; you’ll be asked to share the information with the freelancers anyway. In the “Hire” page, it’s easy to browse through hundreds (if not thousands) of CAD professionals available for freelance projects. You can filter the search based on location, preferred tools or software, and even education. Every freelancer’s profile comes with additional information like the number of completed projects and client reviews as well.

Website: Behance.net

17. Dribbble

It’s another freelancing site built for graphic design professionals. Hiring a freelancer directly from Dribble comes with a basic 7.5% platform fee, but you don’t have to pay until the project is done. All the data and intellectual property are usually included, unless you and the CAD drafting and design freelancer set a separate agreement regarding the matter. The ability to post a job, however, is available only with a premium account, which costs $150 per month. If you want to have a complete hiring suite with premium features, the monthly subscription is $300.

Website: Dribbble.com

18. Wellfound

One thing that makes Wellfound stand out from the crowd is the self-claim about how it’s a freelancing site built specifically for startups. To call it a freelancing site wouldn’t be entirely accurate, because it looks more like a job board than anything else. By focusing on small businesses, Wellfound makes posting a job on the site free of charge. No matter how many times you do it, you don’t have to pay a dime. Wellfound says that every job posting will immediately reach a community of freelancers ready to take on the next opportunity. A typical job post receives thousands of views within the first week, so it really doesn’t sound too bad indeed. It even has its own ATS platform and the option to boost the job listing (for a fee) to gain a larger exposure.

Website: Wellfound.com

19. Crowdspring

In case you haven’t heard, Crowdspring is a freelancing site that focuses on the graphic design industry. There are two methods to find your next freelancer in Crowdspring. The first and most typical method is by launching a contest. It can be a logo, packaging, illustration, visual art, or anything else that you can use for print and online publications. A number of designers submit their works, you choose the best one, and release the payment. Second, Crowdspring offers the possibility to work with a specific designer on a project. You have the option to choose your own freelancers or use those recommended by the platform. All the tools you need to manage the project are available on the platform.

Website: Crowdspring.com

20. Upwork

One of the direct competitors of Guru, the freelancing platform Upwork, is another well-known option to find remote talent. You can find 3D design freelancers by either posting a project or browsing through their profiles directly. Like a lot of other freelancing sites, Upwork takes a percentage of the freelancer’s earnings from every completed project. While it might look like a disadvantage to the freelancers, the reality is that it’s a common practice–the platform simply acts as an intermediary party that connects your business to the talent pool, and it costs money to do so. Upwork offers several tiers of membership for clients, such as Basic, Plus, and Enterprise.

Website: Upwork.com

21. Coroflot

Posting a single project on Coroflot will set you back $295. But the good thing is that the design category is pretty comprehensive, with freelancers looking for their next gigs in concept art, architecture, visual design, product design, 3D modeling, and more. The job posting should remain live for about three months, and you can edit it at any time during that period. Coroflot also offers some bundle packages, such as the 3-pack ($265 per job post), 5-pack ($235 per job post), and 10-pack ($200 per job post). The fee is pretty hefty compared to many other sites, but not unreasonable either.

Website: Coroflot.com

22. Designhill

Claiming to have more than 150,000 freelancers on the network, Designhill sure is a promising place to source freelancers in the graphic design field. You can post a design project as a public contest to get multiple submissions (but you can only use the winning design, of course) or as a one-to-one project with a specific freelancer. Even if you’ve never used a freelancing site before, Designhill has a clear layout on top of a pleasing user interface that you won’t find difficulties navigating the website.

Website: Designhill.com

23. Hubstaff Talent

If you want a bare-bones, no-frills, minimalistic approach to hiring freelancers, Hubstaff Talent is the site you’re looking for. It’s a completely free platform where companies or employers can connect directly with 2D & 3D drafting freelancers from all around the world. There’s no platform fee, no markups, and no middlemen involved in the hiring process. You can set up an alert to get a notification each time someone responds or applies to your project. While Hubstaff Talent does offer a management platform for time tracking, you are under no obligation to use it. When communicating with applicants and freelancers, you’re free to use any channel you like; it can be email, Skype, social media, or anything else you prefer. Hubstaff Talent only provides the space to post a project and make first contact with freelancers.

Website: Hubstafftalent.com

24. Truelancer

The design category in Truelancer is filled with freelancers specializing in various fields, from visual arts to interior design. Truelancer offers two methods to find the right freelancers for your needs: contest and direct hire. The former refers to posting a project for the public to see and receiving design submissions from participants. You don’t get to choose who the participants are, but the submissions should provide you with enough design options. The direct hire option means Truelancer will help you find the right freelancers based on your project brief using AI.

Website: Truelancer.com

25. Virtual Vocations

Every freelancer registered with Virtual Vocations is looking for a remote project. In fact, the platform specializes in work-from-home jobs and provides employers with access to tracking statistics, such as job post views and CTR. As an employer looking for freelancers, you have to register with Virtual Vocations to be able to post jobs. Once your registration is approved, you have the freedom to post an unlimited number of job postings, as long as the job is meant for work-from-home freelancers. The good thing is that freelancers can apply directly by email or company link; they don’t have to communicate with you on the platform.

Website: Virtualvocations.com

26. Easy Render

With a massive network of 3D artists from more than 130 countries worldwide, Easy Render is a promising place to find and hire your next freelancer. Mind you that the platform is specifically built for the architectural visualization sector of the design industry–it’s all about photorealistic 3D rendering of architectural plans such as interior, exterior, furniture pieces, floor plans, etc. Also, it costs nothing to register an account with Easy Render. Posting a job is free as well. You’ll only pay once you approve the artist’s work and the project is done, but the site requires you to deposit the amount into escrow beforehand.

Website: Easyrender.com

27. Treatstock

If your next design project has anything to do with 3D printing services or CNC machining, Treatstock can be a great place to search for freelancers to get the job done. Instead of browsing through hundreds of freelancers on the site, the internal team at Treatstock will provide you with a list of recommended designers based on your project description. Most projects are about creating 3D objects of various shapes and forms. As long as the objects can be produced using a 3D printer or CNC machining process, Treatstock has you covered. When the project is done and you receive the final design, you can have Treatstock produce it for you, too.

Website: Treatstock.com

28. Fiverr

The interesting thing about Fiverr is how the site is mostly built for freelancers, instead of the other way around. Rather than posting your project to the site, you are provided with a simple search bar to tell the site what services you’re looking for, and the site will give you a list of freelancers specializing in the field. You can post a job re quest, but there’s no way to track how many freelancers have “viewed” the post at any given time. Fiverr will match the project brief with some sellers, and it will notify you when one of them accepts the request. It really is just easier to search for CAD design and browse through the freelancers’ profiles and portfolios.

Website: Fiverr.com

29. Workana

Unlike many other freelancing sites, Workana prioritizes workers and job seekers based in Latin America. The design category is filled with dozens of 3D rendering freelancers specializing in CG animations, electronic designs, 3D designs, and more. Signing up is free, and you can post a project right away after you complete the registration process. Interested freelancers may respond to your project with their proposals, so it’s basically a bidding process. You’re allowed to hire the winning bidder for a flat rate or an hourly rate. Either way, you must deposit the payment in advance into an escrow account. The payment is released only when the project concludes.

Website: Workana.com

30. Archionline

While it may sound like a bit of a stretch to say that architects and general contractors are freelancers, there’s no denying that they do work for clients in more or less the same fashion. You hire them to complete a project with a flat rate or an hourly rate payment option. And let’s not forget that the jobs of modern architects and general contractors involve quite a lot of CAD applications. With that in mind, Archionline isn’t exactly out of place in this list. To post a project to the site, you need to fill out a simple form to describe the job in detail. It might be helpful to include a picture or two as references. A project manager from Archionline will reach out to you to clarify the information and point you to a capable architectural design expert or general contractor near the project’s location. Of course, you can negotiate the terms of the proposal further with the project manager to get the best deal. The job starts after you pay the deposit online.

Website: Archionline.com

31. Freelancer

The site says that it connects more than 77 million employers and freelancers from all over the world. It’s like any other freelancing website, but with a little bit more sophistication. When you’re about to post a project, you’ll be provided with a series of simple questions to define the job. It starts with a description of the project all the way to options about NDA agreement, payment schedule (hourly rate or fixed-price), and budget. There are additional options like free or premium posting, with the latter giving you access to some sort of project manager to keep everything organized. As soon as the post is live, you’ll begin to receive bids from freelancers.

Website: Freelancer.com

32. PE4H

Professional Engineers for Hire, styled as PE4H, is a platform where you can post an engineering project, receive proposals from engineering design freelancers, and hire the best candidate for the job. The site claims to be able to connect you with a pool of professional engineers nationwide, and you’ll receive notifications for every proposal. You can use the dashboard to review proposals and communicate with candidates online. Most engineers registered with the site are willing to work on either short-term or long-term projects.

Website: Peforhire.com

33. Blender Artist

Despite the clear name similarity, Blender Artist is an online community that’s not at all affiliated with the open-source 3D CAD software, Blender. Just by looking at the homepage of the site, you’ll immediately get the sense that it’s a specialty online forum, frequented only by those with a great interest in the software or using it for professional work. Users are allowed to post a project and invite freelancers to collaborate on it. Blender Artist is maintained by only a small team of moderators, and yet the site is heavily moderated to maintain content quality. Clients can only post paid jobs, and freelancers are advised against working without a signed contract. Whether you need 3D rigging services, CG animations, models, textures, architectural visualizations, or character designs, Blender Artist has just the right pool of talent for the project.

Website: Blenderartist.org

34. Creativepool

The recruitment platform, Creativepool, offers three different methods of hiring job seekers. The first method is called “studiogigs” and made specifically for freelancing purposes. You can post a project for free, and Creativepool says that all freelancers who respond (by sending proposals or bids) are approved professionals. The studiogigs option comes with zero commission fee, but you manage everything by yourself and pay the freelancers directly. The second method is referred to simply as “standard,” but only for a permanent position. The third option, “Premium Featured” has every feature that comes with standard plus a dedicated account manager. Since you’re looking for freelancers, studiogigs is the obvious choice here.

Website: Creativepool.com

35. Archinect

As the name suggests, Archinect is built only for architecture-related matters. To find a freelance CAD designer on the site, simply use the “Talent Finder” feature and sort the talents based on years of experience, educational background, and portfolio projects. Apart from architecture, other areas of specialization include construction, engineering, landscape, interiors, furniture design professionals, and urban planning. In case you want to post a design contest instead of browsing the profiles directly, use the sister site, Bustler.

Website: Archinect.com

36. A2D

Sometimes you get to find your way to the best freelancers around by running and sponsoring a contest managed by an online platform. A2D is only one of many places where you can do exactly that. To post a contest here, first you have to make sure that the CAD project falls within any of the following categories: mechanical engineering, prototype, industrial design, or concept creation. A design contest gives you the opportunity to see how the participants tackle a customized challenge. And at the end of the day, you’re equipped with more complete information about their skills and level of expertise to make a better, informed hiring decision.

Website: A2d.app

37. Zerply

Everything in Zerply, including the freelancers registered with the site, is all about CG animation and VFX. This isn’t the place where you post a project for an architectural design draft or an engineering design task, but if you want an architectural walkthrough with animation or product visualization in a 360-degree interactive image format, Zerply is right up your alley. To directly hire a freelancer on the site, remember to use the self-serve platform rather than the concierge option. The former allows you to search VFX artists by skills, experience levels, preferred software, location, and availability. Once you find a match, you can contact the freelancer with a few clicks.

Website: Zerply.com

38. Tasker

The best way to describe Tasker is that it’s a managed freelancing service for hardware engineering projects. You start by posting a job with a clear description of the deliverables, budget, and timeframe. Next, Tasker connects you with vetted design engineering candidates who have the skills and experience that match your description. Remember that a hardware engineering project doesn’t always mean building an actual piece of hardware for a mechanical assembly; it can be anything from creating a product concept in 3D format to running a finite element analysis. You pay the freelancers once you approve all the deliverables.

Website: Taskerplatform.com

39. Toptal

Here is the big claim: Toptal says it only has the top 3% of all the freelance talents everywhere, indicating that you’ll be dealing only with the most experienced professional freelance CAD designers, should you decide to use the platform. There’s an account manager to help you with finding/building a team of freelance professionals right from the start, so you basically never have to lift a finger after posting the request. Suppose you only need an individual freelancer rather than a team; there’s an option for that, too. In 2023 alone, Upwork has managed more than 64,000 projects for over 25,000 clients.

Website: Toptal.com

40. ArchitectureQuote

Architects who sell their services through private business entities are basically contractual workers. And all contractual workers are essentially freelancers. This is where ArchitectureQuote comes in. You can hire an architectural rendering designers through the platform in three easy steps. First, you post a project and include additional files, including the early design concept (a sketch will do) as well as technical drawings for builders and engineers. It also requires you to “at least” prove that you’ve already approached the local authority for a permit/approval. Second, upload some inspirational images if needed. Third, the platform matches your project with the right architect. The first step alone seems like a lot of work on your part, but it can only mean that you will receive a more accurate quote.

Website: Architecturequote.com

41. LinkedIn

Everybody has heard of LinkedIn before. It prides itself as a professional network (and claims to be the world’s largest at that too) that brings together companies, employers, and professionals from everywhere in the world in a single online platform. Many professionals on LinkedIn probably wouldn’t consider themselves “freelancers” and prefer to be regarded as “independent contractors” instead. However, independent contractors are technically freelancers, and LinkedIn has no shortage of professionals looking to get hired for short-term projects. In case you’re in the market for professionals in CAD design who are willing to work remotely, LinkedIn’s built-in search function should help you narrow down the search.

Website: Linkedin.com

42. Freelance.com

You probably have heard about Software as a Service and Platform as a Service before, but Freelance.com is saying that it’s doing what’s called Talent as a Service–a fancy way of saying that it connects you with freelancers from various educational backgrounds and experience levels in various industries. There’s no mention that the platform focuses on specific fields, which should only mean it can work with any client working on any project, including CAD design services. The site is based in France, but its network of freelancers is spread across multiple countries like Germany, the UK, Belgium, Switzerland, and Morocco.

Website: Freelance.com

43. DesignContest

The site is exactly what it says in the name. DesignContest is a platform to help you post a design contest, more specifically in the graphic design discipline. Although the contest is the site’s biggest selling point, it has a feature where you can assign a project to an individual designer (or a small team of them) in a one-on-one environment. You write your own design brief and select a few designers to work on it. The one-on-one option is best if you’re confident enough about the freelancers’ abilities or have actually worked with them before.

Website: Designcontest.com

44. Twine

From CG animators to product designers, Twine has all the professional freelancers you need for every graphic CAD project. The “Standard” subscription package gives you free-of-charge job posting and automatic AI-powered candidate screening. You only pay a small platform fee when you actually end up hiring a freelancer through the site. The service fee starts from 5%, but it gradually goes down as you post more projects and hire more freelancers. The “Business” package comes with a $139.99 premium, and for that, you get all the features from the Standard option, added with a manual vetting process and a dedicated account manager.

Website: Twine.net

45. YunoJuno

You won’t be able to hire freelancers directly here. YunoJuno kind of asks you to use the platform to “book” freelancers and manage the project, too. While you can search for 2D & 3D modeling professionals using the search feature, there’s no way to get around the system to contact the freelancers by email, Skype, or social media–you have to use YunoJuno for that. Once you become a registered member, you get access to tens of thousands of freelancers specializing in various fields, including CAD designs.

Website: Yunojuno.com

46. PeoplePerHour

Millions of freelancers from all around the world, including those in the 3D CAD industry, are looking for their next projects on PeoplePerHour. To look for freelancers, simply use the search bar on the homepage; it just doesn’t get any simpler than that, although apparently there is now an AI feature to assist you with the hunt. Once you’re on the search results page, it’s easy to filter the results based on the expected delivery time, the freelancer’s country, and the budget. Mind you that PeoplePerHour does very little in terms of evaluating the members, so you need to manage the search and take on some quality control measures yourself.

Website: Peopleperhour.com

47. Remote OK

Of all the freelancing sites mentioned in this list, Remote OK is probably the most straightforward platform of them all. It’s as raw as it can possibly get for a job board, where you post a project and wait for proposals from freelancers. It doesn’t have a built-in ATS of any sort, which is probably a good thing, since you can communicate with freelancers directly. The not-so-good thing about such a system (basically an online job board) is that you must do your own vetting.

Website: Remoteok.com

48. Remote

The similarly named Remote is an entirely different site. It’s a subscription-based freelancing site with a starting membership fee of $119 per month. According to the platform itself, every project is visible to a pretty massive audience because the site has around 2 million monthly visitors. Remote does not in any way fiddle with the job posting, meaning it offers no premium feature to boost visibility or highlight any project submitted to the site.

Website: Remote.com

49. DeviantArt

Built primarily as an online community for digital artists, DeviantArt has more than 90 million registered members from all around the world. If you head to the “Employment Opportunities” page of the Forum section, you should see two main threads: Artists for Hire and Hiring an Artist. The former is where digital artists promote their skills and services to employers, whereas the latter is for employers to post projects and look for freelancers. You can actually use both threads to browse through freelancers’ profiles and hire one of them to work on your next graphic design project.

Website: Deviantart.com

50. We Work Remotely

To post a job to WWR, you must be a premium member. Subscription fee starts at $299, with some optional upgrades available. The site is intended for employers and freelancers who search specifically for remote hiring opportunities. Since this is a general freelancing site, there’s no limitation on the type of project you can post. WWR says that all applicants have been manually vetted, so the freelancers responding to your job post are regarded as the “priorities” to stop you from wasting time sorting dozens of profiles.

Website: Weworkremotely.com

How Cad Crowd can help

Hiring freelancers from Cad Crowd offers flexibility, cost-efficiency, and access to a global talent pool, making it an attractive option for businesses of all sizes. Unlike full-time employees, freelancers can be hired for short-term or specific projects.

Companies hire remote freelancers to access specialized skills without long-term commitments, gain flexibility in staffing based on project needs, reduce costs by avoiding additional overhead, and tap into a global talent pool, allowing them to find the best fit for specific projects regardless of location. In essence, hiring freelancers provides greater business agility and adaptability to workforce demands while saving money. Contact us for a quote.

MacKenzie Brown is the founder and CEO of Cad Crowd. With over 18 years of experience in launching and scaling platforms specializing in CAD services, product design, manufacturing, hardware, and software development, MacKenzie is a recognized authority in the engineering industry. Under his leadership, Cad Crowd serves esteemed clients like NASA, JPL, the U.S. Navy, and Fortune 500 companies, empowering innovators with access to high-quality design and engineering talent.

Connect with me: LinkedIn ✦ X ✦ Cad Crowd

NVIDIA GTC 2026: Live Updates on What’s Next in AI

Posted on March 20, 2026 by faz_business

Wednesday, March 18, 5:30 p.m. PT 🔗

20 Years of CUDA: Honoring the Architects of the Accelerated Age

What began in 2006 as a bold parallel computing bet has evolved into the foundational heartbeat of modern science and AI.

At GTC, NVIDIA is marking two decades of CUDA — representing the efforts of over 6 million developers innovating across every layer of the computing stack. Today, it serves as a generational bridge between the pioneers who wrote the first kernels and the next wave of builders deploying trillion-parameter AI models.

Led by NVIDIA CUDA Architect Stephen Jones, a panel at GTC Wednesday featured a group of researchers and engineers from Jump Trading, Meta Superintelligence Labs and NVIDIA who highlighted the decades of innovation behind CUDA, how it helps developers solve some of the world’s most complex problems — and how systems like the NVIDIA DGX Spark desktop AI supercomputer will enable the next generation of CUDA developers.

The group shared memories of the early days of CUDA — when “nobody wanted GPUs,” said Paulius Micikevicius, a software engineer at Meta Superintelligence Labs. “We had to go and beg them to consider using GPUs.”

During that time, Wen-Mei Hwu, senior distinguished research scientist and senior research director at NVIDIA, then a professor at the University of Illinois Urbana-Champaign, decided to build a 200-GPU system in two months with a group of grad students.

“A couple of weeks later, 200 GPU boards arrived, and power supply and everything — but there’s no chassis. So we ended up building wood frames for each of these boards … and we ran the Green500 [benchmark] and we got No. 3,” Hwu said. “That was the moment I realized that the energy efficiency of GPUs has incredible potential.”

As the scale of accelerated computing has shifted to rack-scale systems and AI factories, the panelists see desktop AI systems like DGX Spark as a new way forward for prototyping and early development.

“As long as you have that capability to do that initial exploration and something that fits on your desk or your lap, that’s the critical thing,” said Kate Clark, distinguished devtech engineer at NVIDIA. “I don’t see that going anywhere anytime soon. We’ll always have CUDA everywhere.”

Monday, March 16, 1:30 p.m. PT 🔗

NVIDIA cuDF and cuVS Adopted by World’s Leading Data Platforms, Fueling Modern Enterprise Data Processing

Enterprises are generating hundreds of zettabytes each year, and organizations are racing to turn that information into insights. NVIDIA cuDF and cuVS — accelerated data libraries built on NVIDIA CUDA‑X — are being adopted by data platforms across industries to deliver up to 5x faster performance while reducing costs for structured and unstructured data processing.

Integrated with the world’s most widely used open source data engines — downloaded over 200 million times monthly by developers — these libraries are harnessed across enterprise data platforms, databases and data lakes. This helps organizations accelerate innovation, develop more accurate models and process more data while managing costs.

For structured data, NVIDIA cuDF accelerates open source data processing engines such as Apache Spark, Presto, DuckDB, Polars and Velox, delivering up to 5x faster processing compared with CPU-only deployments.

For unstructured data — which represents 80% of today’s enterprise data and is growing rapidly — NVIDIA cuVS accelerates leading engines including FAISS, Amazon OpenSearch Service and Milvus. This helps agents and applications extract context, facts and recommendations from vast stores of text, images and video in a fraction of the time.

Powering Enterprise Data Processing Platforms

Google Cloud integrates NVIDIA cuDF to accelerate Apache Spark within Dataproc and cuDF can be easily used within Google Kubernetes Engine (GKE) to reduce processing times for massive ETL jobs from hours to seconds while lowering compute costs.

At Snap, which serves more than 946 million active users, NVIDIA cuDF on GKE cut daily data processing costs by 76%. This enables 10 petabytes of data to be analyzed within a three-hour window — saving millions of dollars.

“Our collaboration with NVIDIA and Google Cloud helps us innovate faster for more than a billion Snapchatters worldwide,” said Saral Jain, chief information officer of Snap. “By lowering data processing costs and scaling experiments across petabytes of data, we’re delivering AI-powered experiences more quickly and efficiently.”

IBM watsonx.data is a hybrid, open data platform that includes open source analytics engines such as Apache Spark and Presto engines for structured data, and a vector engine based on OpenSearch. In early experiments with Nestlé’s Order-to-Cash mart, watsonx.data with NVIDIA cuDF accelerated workloads ran five times faster, with 83% lower cost savings.

“For a company that serves billions, data underpins decision making across our global operations,” said Chris Wright, chief information and digital officer of Nestlé. “Working with IBM and NVIDIA, a targeted proof of concept has demonstrated the ability to refresh global operations data in a few minutes and at reduced cost. Our focus now is on turning this capability into tangible business impact — further improving decision speed in areas such as manufacturing and warehousing, and scaling these capabilities across our enterprise.”

The Dell AI Data Platform with NVIDIA includes accelerated data engines that enable enterprises to quickly and securely activate their Dell AI Factory with AI-ready data. It features an Apache Spark-based processing engine accelerated with NVIDIA cuDF, delivering up to 3x faster performance, and an enterprise-grade vector database accelerated with NVIDIA cuVS, delivering up to 12x higher throughput for vector indexing compared with CPUs.

“Purpose-built for agentic AI, the Dell AI Data Platform with NVIDIA uses accelerated data processing engines to make multimodal data AI-ready in hours instead of days,” said Michael Dell, chairman and CEO of Dell Technologies.

Oracle announced that Oracle Private AI Services Container can greatly accelerate vector index creation in Oracle AI Database using NVIDIA cuVS, helping organizations speed up AI-enabled decisions with the latest information.

“Enterprise AI is moving from experimentation to production,” said Clay Magouyrk, CEO of Oracle. “Oracle AI Database with NVIDIA technology delivers AI-ready data within minutes, enabling applications that were previously impossible.”

NVIDIA cuDF and cuVS are supported by leading enterprise data platforms including EDB Postgres AI, NetApp, Snowflake, Starburst and VAST Data — setting the foundation for the AI‑powered future of data processing.

Monday, March 16, 1:30 p.m. PT 🔗

NVIDIA Launches cuEST for Accelerated Quantum Chemistry in Semiconductor Design

NVIDIA this week launched NVIDIA cuEST, a new NVIDIA CUDA-X library that shifts electronic-structure calculations onto GPUs. Applied Materials, Samsung, Synopsys and TSMC are among the initial adopters.

A leading-edge chip now contains over 50 billion transistors. Engineering them requires answering fundamental physics questions at the atomic scale: how electrons bond, how they migrate and how they interact across films just a few atoms thick.

“As semiconductor scaling reaches the physical limits of materials, the industry requires a massive increase in computing performance to simulate the quantum mechanics of next-generation chip designs,” said Tim Costa, general manager for industrial and computational engineering at NVIDIA. “With NVIDIA cuEST, industry leaders can move past the quantum bottleneck and take high-fidelity chemical modeling directly into production to accelerate semiconductor innovation.”

Industry Impact

Applied Materials: Applied Materials uses cuEST-accelerated density functional theory (DFT) to model challenging structures, predict material properties and study reaction pathways.
Samsung: Samsung integrated cuEST into its internal pipeline, already accelerated on GPUs, to deliver yet another up to 5x end-to-end speedup for key quantum-chemistry workloads.
Synopsys: Powered by cuEST and QuantumATK, Synopsys expanded its functionality to include Gaussian-basis DFT, accelerating simulations up to 30x for semiconductor workflows.
TSMC: TSMC uses cuEST’s accelerated quantum chemistry to advance processes for next-generation silicon design.

From the Lab to the Fab

The most common method for atomistic modeling is density functional theory. DFT offers a strong balance between accuracy and scalability; however, its computational cost has limited its widespread use in industry, keeping most applications confined to research. With cuEST, NVIDIA makes high‑accuracy quantum‑chemistry feasible at an industrial scale and in real production workflows.

Historically, the industry has relied on CPU clusters to run these simulations, evaluating candidate materials, including gate dielectrics and interconnect metals, one batch at a time over hours or days.

cuEST provides optimized routines so GPUs can accelerate the core matrices of a Gaussian-basis DFT calculation, including overlap, kinetic energy, nuclear attraction, Coulomb and exchange-correlation. It also supports functional approximations ranging from standard generalized gradient approximation to hybrid functionals, allowing engineers to balance computational cost with accuracy.

NVIDIA’s goal for cuEST: moving high-fidelity material modeling from the lab to the fab.

Learn more about cuEST by joining the NVIDIA demo booth and Synopsys’ booth at GTC, and dive deeper in the GTC session, “Next-Generation Discovery: Agentic AI for Science, AI-Driven Simulation and GPU-Accelerated Chemistry.”

Amazon acquires Rivr, maker of a stair-climbing delivery robot

Posted on March 20, 2026 by faz_business

Rivr, a Zurich-based autonomous robotics startup known for its stair-climbing delivery robot, has been acquired by Amazon in a deal that signals the e-commerce giant’s interest in doorstep delivery. Terms of the deal weren’t disclosed.

Co-founder and CEO Marko Bjelonic, who once described the four-legged wheeled robot to TechCrunch as a “dog on roller skates,” shared the acquisition news on LinkedIn. The Information was first to report the deal.

Bjelonic said in his LinkedIn post that the acquisition will “accelerate our vision of building General Physical AI through doorstep delivery, bringing robotics and AI closer to real-world deployment at scale,” meaning, in plain terms, that Amazon’s resources should help Rivr get its robots onto more doorsteps, faster.

Last year, Rivr launched a pilot program in Austin with Veho, a package delivery company. Bjelonic said, at the time, he hoped to learn from the partnership with Veho and eventually scale to 100 bots by 2026. It’s unclear if the company was ever able to reach that milestone.

TechCrunch has reached out to Rivr for comment.

Rivr got the attention of Amazon long before its pilot program. The Amazon Industrial Innovation Fund and Bezos Expeditions invested in Rivr as part of a $22.2 million seed round that closed in 2024, according to PitchBook. The startup, which had raised a total of $25 million, was last valued at $100 million.

Techcrunch event

San Francisco, CA
|
October 13-15, 2026

Perplexity Can Now Access Your Apple Health Data to Answer Medical Questions

Posted on March 20, 2026 by faz_business

AI company Perplexity today introduced Perplexity Health, a suite of connectors that allow Perplexity to access your health data.

perplexity health
Perplexity suggests that the feature is useful for aggregating health data from across different portals, apps, and devices. It is able to track metrics and trends over time across biomarkers, with information shown on a personalized dashboard.

When asked a health-related question, Perplexity Health can answer based on medical records, lab results, and wearable data. With Perplexity Computer (Perplexity’s AI agent tool), AI agents can use health information to build personalized fitness plans, nutrition plans, and more. Perplexity Health on Computer is rolling out to Pro and Max users in the U.S. first.

Perplexity says that Perplexity Health draws from “premium medical literature” like clinical guidelines and peer-reviewed journals. Perplexity has established a Perplexity Health Advisory Board with physicians, researchers, and health tech experts who will “pressure-test” product decisions, content quality, and clinical safeguards against evidence-based medicine standards.

Perplexity Health is able to connect to Apple Health, so it can integrate data collected by the Apple Watch and added to the Apple Health app. It also supports Fitbit, Ultrahuman, and Withings, along with electronic health records from more than 1.7 million care providers. Oura and Function integrations are coming soon.
Health data is encrypted, and Perplexity says there are strict access controls and tools to manage or delete information at any time. Health information is not used to train AI models or sold to third parties.

Perplexity is the second AI company to integrate with Apple Health. OpenAI introduced a ChatGPT Health feature with Apple Health support in early 2026.

Wonders of Sodor Free Download (Build 22317314+1 DLC)

Posted on March 20, 2026 by faz_business

Thomas and Friends Wonders of Sodor Preinstalled Worldofpcgames

Thomas and Friends: Wonders of Sodor Direct Download

Join Thomas and his friends on a special journey across Sodor! Hop into the cabs of North Western Railway’s classic steam team to enjoy nostalgic stories and all-new narratives right from the drivers’ seats. Explore the Wonders of Sodor – fun and magic await!

Take the controls of Thomas, Percy, Gordon, Emily, and Diesel to explore iconic landmarks across the Island of Sodor. At Sir Topham Hatt’s invitation, carry passengers in Annie & Clarabel along Thomas’ branch line, hurry express passengers over the top of Gordon’s Hill, or transport freight with the Troublesome Trucks.

Thomas & Friends: Wonders of Sodor tells classic and original tales inspired by the Railway Series books and TV shows, now told for the first time in an interactive, first-person perspective – the first ever faithful recreation of Sodor in a game! Skeleseller

Experience the sights, sounds, and feelings of Sodor as you explore iconic Thomas & Friends locations like you’ve never seen them before, including Tidmouth Sheds, Knapford station, and the famous Windmills. A total of 8 stories await including Thomas & Gordon, Thomas & Bertie’s Great Race, Percy’s Troublesome Day, and more. Each story features full narration by the legendary Mark Moraghan for the ultimate storytelling immersion.

But that’s not all! Help Thomas and his Friends be really useful engines across a full working schedule for the Island of Sodor. See all the engines running about the island doing their daily tasks, get involved by driving Thomas on his branch line, or even ride as a passenger! While exploring the island, look out for hidden collectables to find!

Features and System Requirements:

Join Thomas and friends on a cheerful journey across the Island of Sodor filled with fun activities.
Drive trains and complete tasks while helping characters and delivering important cargo.
Explore colorful locations that bring the world of Thomas & Friends to life.
Take part in simple challenges designed for easy and enjoyable gameplay for all ages.

Screenshots

System Requirements

Minimum
Requires a 64-bit processor and operating system
OS: 64-bit Windows 10 or Windows 11
Processor: Intel Core i5-4690 @ 3.5 GHz or AMD Ryzen 5 1500X @ 3.7 GHz
Memory: 8 GB RAM
Graphics: NVIDIA GeForce GTX 750 Ti or AMD Radeon RX 460 with 2 GB VRAM or more
DirectX: Version 10
Storage: 35 GB available space
Support the game developers by purchasing the game on Steam

Installation Guide

Turn Off Your Antivirus Before Installing Any Game

1 :: Download Game
2 :: Extract Game
3 :: Launch The Game
4 :: Have Fun 🙂

Reducing GPU Memory and Accelerating Transformers

Posted on March 20, 2026 by faz_business

Introduction

The transformer revolution is now deep into its long‑context era. Models like GPT‑4 (32 k tokens), MosaicML’s MPT (65 k), and Claude (100 k) can process entire chapters or codebases. Yet as context grows, the attention mechanism becomes the bottleneck: calculating the similarity matrix S = Q·K^T and the probability matrix P = softmax(S) produces N×N data structures. These matrices must be moved between the GPU’s tiny on‑chip SRAM and its larger but slower high‑bandwidth memory (HBM), consuming bandwidth and limiting throughput. In a world where compute FLOPs continue to climb, the real constraint has become memory.

FlashAttention, introduced in 2022, addressed this problem by tiling the computation to avoid ever storing the full S or P matrices, delivering 2–4× speedups and up to 10–20× memory savings. FlashAttention‑2 (FA2) goes further: it reduces costly non‑matmul operations, parallelizes across sequence length, and partitions work to minimize shared‑memory traffic. Benchmarks show FA2 is about twice as fast as its predecessor and up to nine times faster than standard attention implementations, hitting 225 TFLOPs/s on NVIDIA A100 GPUs. This guide explains how FA2 works, when to use it, how to integrate it into your stack, and where its limits lie.

Quick Digest

FA2 solves a memory‑bound problem. Attention’s N² memory footprint stalls GPUs; tiling and kernel fusion bring it down to linear memory cost.
Key innovations: fewer non‑matmul FLOPs, extra parallelism along sequence length, and slicing the query matrix across warps.
Adoption: Supports Ampere/Ada/Hopper GPUs and FP16/BF16 datatypes. Install via pip and flip a flag in PyTorch or Hugging Face to enable.
Who benefits: Anyone training or serving long‑context models (8 k–16 k tokens) or using large head dimensions; cost savings are substantial.
Caveats: Only attention is accelerated; feed‑forward layers remain unchanged. FP32 precision and older GPUs are unsupported.

The Memory Bottleneck in Transformers

Why memory—not compute—matters

Each token attends to every other token, so naïve attention materializes N×N matrices. With 4 k tokens and 96 heads, the similarity and probability matrices alone consume several gigabytes. On modern GPUs, data movement between the tiny on‑chip SRAM (≈20 MB) and HBM (≈40–80 GB) dominates runtime. More compute doesn’t help if the algorithm shuttles large intermediate results back and forth.

To decide whether you need FA2, perform the MEMS Check:

Memory – Estimate your attention matrix size. If it can’t fit in SRAM and triggers out‑of‑memory errors, you’re memory‑bound.
Efficiency – Use profilers (Nsight or PyTorch) to see if kernels saturate compute or stall on memory transfers.
Model size – Many heads or large embeddings increase memory overhead.
Sequence length – Beyond ~2 k tokens, standard attention’s O(N²) memory explodes.

If two or more factors flag red, FA2 can help. However, tasks with short sequences (≤512 tokens) remain compute‑bound and won’t benefit from tiling; the overhead of custom kernels may even slow them down.

Expert insight

“FlashAttention exploits the asymmetric GPU memory hierarchy to bring significant memory saving and 2–4× speedups without approximation.” – Dao et al.

Understanding that memory—not computation—limits attention is key to appreciating FA2’s value.

Quick summary

Why does memory limit attention? Because attention creates huge N² matrices that must be moved between slow and fast memory. Profilers help determine if your workload is memory‑bound.

FlashAttention Fundamentals—Tiling and Recomputing

Tiling and kernel fusion

FlashAttention reorders computation to avoid ever materializing the full N×N matrices. It divides queries (Q), keys (K), and values (V) into blocks that fit in SRAM, performs matrix multiplications and softmax operations on those blocks, and accumulates partial sums until the final output is produced. Because all intermediate work stays on‑chip, memory traffic drops dramatically.

Kernel fusion plays a crucial role: instead of launching separate CUDA kernels for matmul, scaling, softmax, masking, dropout, and value projection, FlashAttention performs them within a single kernel. This ensures that data isn’t written back to HBM between steps.

Recomputation in the backward pass

During backpropagation, naïve attention must store the entire attention matrix to compute gradients. FlashAttention saves memory by recomputing the necessary local softmax values on the fly. The small cost of extra computation is outweighed by eliminating gigabytes of storage.

Negative knowledge

FlashAttention doesn’t alter the mathematical formula for attention; any deviations in output typically arise from using lower precision (FP16/BF16). Early versions lacked dropout support, so ensure your library version accommodates dropout if needed.

Quick summary

How does FlashAttention reduce memory? By tiling Q/K/V into blocks, fusing operations into a single kernel, and recomputing softmax values during backprop.

What’s New in FlashAttention‑2

FA2 refines FlashAttention in three major ways:

Fewer non‑matmul operations: GPUs achieve enormous throughput on matrix multiplication but slow down on general FP32 operations. FA2 rewrites rescaling and masking code to minimize these non‑matmul FLOPs.
Parallelism along the sequence dimension: When batch size × head count is small, the original FlashAttention can’t saturate all GPU streaming multiprocessors. FA2 parallelizes across long sequences, boosting occupancy.
Query slicing: Instead of slicing keys and values across warps (requiring synchronization), FA2 slices the query matrix, allowing warps to compute their output independently. This eliminates shared‑memory writes and delivers more speed.

FA2 also supports head dimensions up to 256, as well as multi‑query (MQA) and grouped‑query (GQA) attention. Head dimension support matters for code‑oriented models like CodeGen or GPT‑J.

Decision guidance

Use this quick decision tree:

If you run on Turing GPUs (e.g., T4) –> stick to FlashAttention 1 or standard kernels.
Else if your head dimension >128 –> choose FA2.
Else if (batch_size × num_heads) is small and sequence is long –> FA2’s extra parallelism pays off.
Else benchmark FA1 and FA2; the simpler implementation may suffice.

Caveats

FA2 requires Ampere, Ada, or Hopper GPUs and currently supports only FP16/BF16 datatypes. Compilation is more complex, and unsupported GPUs will fall back to FA1 or standard attention.

Expert insight

“FlashAttention‑2 is about 2× faster than FlashAttention and reaches up to 230 TFLOPs/s on A100 GPUs.” – Tri Dao

FA2 closes much of the gap between attention kernels and optimized matrix multiplications.

Quick summary

What distinguishes FA2? It cuts non‑matmul operations, parallelizes over sequence length, slices queries instead of keys/values, and supports larger head sizes and MQA/GQA.

Installing and Integrating FlashAttention‑2

Requirements and installation

FA2 supports A100, H100, RTX 3090/4090, and AMD MI200/MI300 GPUs and requires FP16/BF16 precision. Install via:

pip install flash-attn --no-build-isolation

Ensure CUDA ≥12.0 (or ROCm ≥6.0) and PyTorch ≥2.2. Install the ninja build system to shorten compile times; if your machine has limited RAM, cap parallel jobs using MAX_JOBS=4.

Enabling FA2 in frameworks

In Hugging Face Transformers, set the use_flash_attn_2=True flag when instantiating your model. For custom code, import and call the kernel:

from flash_attn_interface import flash_attn_func
output = flash_attn_func(q, k, v, causal=True)

Input tensors should be shaped [batch, seq_len, num_heads, head_dim] or as required by the library. For unsupported hardware, implement a try/except block to fall back to standard attention.

Operational advice

GPU orchestration: Platforms like Clarifai’s compute orchestration make it easy to run FA2 on clusters. Select A100 or H100 GPUs, and use the built‑in profiling tools to monitor tokens per second. If you need turnkey hardware, Clarifai’s GPU hosting provides managed A100/H100 instances that integrate with local runners and remote orchestration.
Mixed precision: Combine FA2 with automatic mixed precision (AMP) to maximize throughput.
Benchmarking: After integration, measure tokens per second, GPU memory usage, and wall‑clock time with and without FA2. Use these numbers to adjust batch sizes and sequence lengths.

Quick summary

How do I use FA2? Install the package, ensure you have compatible GPUs and drivers, enable FA2 in your framework, and benchmark. Use Clarifai’s orchestration and model inference tools for scalable deployment.

Performance Benchmarks and Cost Savings

Speedups on A100 and H100

Public benchmarks report that FA2 delivers around 2× speedup over FA1 and up to 9× over standard PyTorch attention. When training GPT‑style models end‑to‑end, FA2 achieves 225 TFLOPs/s on A100 GPUs and even higher throughput on H100 due to newer tensor cores.

An evaluation by Lambda Labs shows that FA2 increases the affordable batch size from 1 to 4 while keeping GPU memory constant; tokens per second jump from 3,717 to 10,650 on A100 and from 6,267 to 22,282 on H100.

Config	Tokens/sec	Batch size	Notes
A100 baseline	3,717	1	Standard attention
A100 FA2	10,650	4	2.9× throughput increase
H100 baseline	6,267	1	Standard attention
H100 FA2	22,282	4	3.5× throughput increase

Scaling to multi‑GPU clusters yields near‑linear performance when high‑bandwidth interconnects (NVLink/NVSwitch) are available.

Cost impact

Because FA2 allows larger batch sizes and higher throughput, it reduces training time and compute cost. For example, replicating GPT3‑175B training with FA2 on 1,024 H100 GPUs is estimated to cost around $458 k, a 90 % reduction compared with traditional kernels. On cloud platforms like Clarifai, fewer GPU hours translate directly into cost savings.

Caveats

Iter/sec may drop slightly because each batch is larger. Actual tokens/sec is the meaningful metric; ensure you measure the right quantity. Multi‑GPU gains depend on interconnect bandwidth; low‑bandwidth clusters may not realize full speedups.

Quick summary

How much faster is FA2? Roughly twice as fast as FA1 and up to nine times faster than standard attention. It increases batch size and reduces training costs dramatically.

Practical Use Cases and Decision Guide

Long‑context language models

FA2 shines when you need to process long documents, stories, or transcripts. With its linear memory cost, you can train or fine‑tune models on 16 k–64 k tokens without approximations. Legal document review, novel writing, and research paper summarization all benefit. Clarifai’s model inference pipeline makes it easy to deploy these large models and serve predictions at scale.

Code and multimodal generation

Models like CodeGen or Stable Diffusion 1.x use large head dimensions (up to 256), which FA2 supports. This allows for deeper code context or higher resolution images without running out of memory.

High‑throughput inference with MQA/GQA

FA2’s support for multi‑query and grouped‑query attention reduces KV cache size and speeds up inference. This is ideal for chatbots and real‑time assistants serving thousands of users concurrently.

Decision matrix

Scenario	Sequence length	Head dim	GPU	Recommendation
Short text classification	≤2 k	≤64	Any	Standard/FA1
Long doc summarization	8 k–16 k	≤128	A100/H100	FA2
Code generation	4 k–8 k	256	A100/H100	FA2
Real‑time inference	≤4 k	≤128	A100/H100	FA2 with MQA/GQA
Ultra‑long context (≥64 k)	>64 k	any	Mixed GPU/CPU	Sparse/approximate

Common mistakes and tips

Don’t assume that bigger batches always improve training; you may need to retune learning rates. Multi‑GPU speedups depend on interconnect bandwidth; check whether your cluster uses NVLink. Finally, remember that FA2 accelerates self‑attention only—feed‑forward layers may still dominate runtime.

Quick summary

Who should use FA2? Practitioners working with long contexts, large head sizes, or high‑throughput inference. Short sequences or unsupported GPUs may not benefit.

Limitations and Alternatives

Precision and hardware constraints

FA2 runs only on Ampere/Ada/Hopper GPUs and AMD’s MI200/MI300 series and supports FP16/BF16 datatypes. FP32 precision and older GPUs require falling back to FA1 or standard attention. Edge devices and mobile GPUs are generally unsupported.

Where FA2 won’t help

If your sequences are short (≤512 tokens) or your model has few heads, the overhead of FA2 may outweigh its benefits. It does not accelerate feed‑forward layers, convolutional operations, or embedding lookups; for these, consider other optimizations.

Alternatives

For extremely long sequences (>64 k tokens) or hardware without FA2 support, consider Performer, Linformer, Longformer, or Paged Attention. These methods approximate attention by using low‑rank projections or local sparsity. They may sacrifice some accuracy but can handle contexts that FA2 cannot.

Quick summary

When should you avoid FA2? When precision must be FP32, when running on unsupported GPUs, when contexts are short, or when approximations suffice for extreme lengths.

Looking Ahead

Emerging kernels

FlashAttention‑3 (FA3) targets the H100 GPU, adds FP8 support, and leverages Tensor Memory Accelerator hardware, pushing throughput even higher. FlashAttention‑4 (FA4) is being rewritten in CuTeDSL for Hopper and Blackwell GPUs, with plans for unified kernels and full FP8 support. These kernels are in beta; adoption will depend on hardware availability.

New attention variants

Researchers are combining hardware‑aware kernels like FA2 with algorithmic innovations. Flash‑Decoding accelerates autoregressive inference by caching partial results. Paged Attention breaks sequences into pages for memory‑efficient inference, enabling 64 k contexts and beyond. FastAttention adapts FA kernels to NPUs and low‑resource GPUs. Expect hybrid techniques that unify tiling, sparsity, and new precisions.

Preparing for the future

To stay ahead, follow these steps: subscribe to flash-attn release notes, test FP8 workflows if your models tolerate lower precision, plan for A100/H100/B200 upgrades, and explore combining FA kernels with sparse attention for ultra‑long contexts. Clarifai’s roadmap includes support for new GPUs and FP8, helping teams adopt these innovations without overhauling infrastructure.

Quick summary

What’s next? FA3 and FA4 target new GPUs and FP8, while variants like Flash‑Decoding and Paged Attention tackle inference and extremely long contexts. Hybrid methods will continue to push transformer efficiency.

FAQs

Q: Does FlashAttention‑2 change the attention computation?
A: No. FA2 preserves the exact softmax attention formula. Differences in output arise from lower precision; use FP16/BF16 accordingly.

Q: Does FA2 support dropout and cross‑attention?
A: Recent versions support dropout and are being extended to cross‑attention. Check your library’s documentation for specifics.

Q: Can I use FA2 with LoRA or quantization?
A: Yes. FA2 operates at the kernel level and is compatible with techniques like LoRA and quantization, making it a good complement to other memory‑saving methods.

Q: What about JAX or TensorFlow?
A: Official FA2 kernels are available for PyTorch. Third‑party ports exist for other frameworks but may lag behind in performance and features.

Conclusion

As transformer models stretch into the tens of thousands of tokens, memory, not compute, is the bottleneck. FlashAttention‑2 provides a timely solution: by tiling computations, fusing kernels, reducing non‑matmul operations, and parallelizing across sequence length, it brings attention performance closer to the efficiency of optimized matrix multiplication. It doubles the speed of its predecessor and dramatically cuts memory use. Real‑world benchmarks confirm that FA2 offers substantial throughput gains and cost savings.

FA2 is not universal; it requires modern GPUs and supports only FP16/BF16. For ultra‑long sequences or unsupported hardware, approximate attention methods remain important alternatives. Yet for the majority of long‑context workloads today, FA2 is the most efficient exact attention kernel available.

Implementing FA2 is straightforward: install the library, enable it in your framework, and profile performance. Platforms like Clarifai’s compute orchestration and model inference simplify deployment across clusters, allowing you to focus on model design and application logic. If you don’t have GPU hardware, Clarifai’s GPU hosting offers ready‑to‑run clusters. And to test these capabilities risk‑free, start for free and claim credits via Clarifai’s sign‑up. Use our MEMS Check to decide whether your workload is memory‑bound, and keep an eye on emerging kernels like FA3/4 and Paged Attention.

In 2026 and beyond, transformer efficiency will hinge on pairing algorithmic innovations with hardware‑aware kernels. FA2 offers a glimpse into that future—one where memory bottlenecks no longer constrain the horizons of our models.

Counter-Strike 2 introduces the game’s biggest shake-up in years: Now when reloading, you lose all the mag’s remaining ammo

Posted on March 19, 2026 by faz_business

Counter-Strike 2 has released an update that makes one of the biggest changes in the game’s modern history: and no-one saw this one coming. Valve has changed how reloading works, so that when you reload, any ammo remaining in your current magazine is lost. And players are not happy.

“When you reload in CS2, the leftover ammo in your magazine is dumped back into an essentially endless reserve supply,” says Valve’s blog post on the change. “And so the decision to reload has never offered significant trade-offs—in a safe position with enough time, you might reload after firing a single bullet, or half a mag, or after firing down to empty, and the rest of the round would be unaffected.

Article continues below

There’s also now a new HUD element, a small bar under the bullet count that empties when you shoot: so it’s easier to tell at a glance roughly how much ammo you’ve got left. You also have a new mag count.

I went through all the Weapon Ammo Changes and here’s all the data that you need ‼️🔼 Biggest Buffs:- Galil AR (+50 Bullets)- M4A4 (+30 Bullets)- P250 (+13 Bullets)- CZ75-A (+12 Bullets)- FAMAS (+10 Bullets)- PP-Bizon (+8 Bullets)- R8 (+8 Bullets)🔻 Biggest Nerfs:-…March 18, 2026

The general trend is more ammo for most rifles, with the P250, CZ75-A and R8 pistols also getting a buffed bullet count. But these go hand-in-hand with some major nerfs, and the two that immediately jump out are the AWP and M4A1-S each having 20 bullets removed. I can see the reasoning behind the AWP: it’s a one-hit kill precision weapon, so if you can’t kill five players with a maximum of 15 bullets maybe the issue isn’t the ammo count.

But even before this reloading change the M4A1-S was one of the weapons where you had to be careful with the ammo, and can easily run out. Again you can argue it’s supposed to be a precision weapon, but I suspect this is one nerf that will be reverted before too long. In total seven weapons have buffed ammo counts, 16 have lost ammo to varying degrees, and 12 guns remain unchanged.

The community reaction across Steam, X and things like the CS2 subreddit is almost universally negative, with a few brave souls putting their head above the parapet to say this might be neat, actually. But generally players are both baffled as to why Valve has made such a fundamental change to Counter-Strike’s mechanics after 27 years, and demanding an immediate reversion of the change (good luck with that).

This graph of the CS2 Steam reviews over the last 24 hours kind of say it all:

A graph showing the Counter-Strike 2 Steam reviews going negative after the controversial reloading update.

(Image credit: Valve Software)

“As a chronic reloader I’m terrified” says reimelcracker on the CS2 subreddit, with Kaka-carrot-cake agreeing that “we are fucked lmao.” One of the few positive perspectives on the change comes from Amnizu in the same thread, who says “I’m fucked too but I love this change. Now it’s more skillful and ammo conservation is actually another thing you need to learn and be aware of. Really excited to see the change in professional games.”

Valve will surely be keeping a very close eye on how things play out: Counter-Strike is one of the most perfectly balanced competitive shooters ever made, and throughout its various iterations the basic core has stayed the same for a reason.

That’s the headline news for this update, though there’s also another significant change, and one I suspect will prove a little more popular. Valve has incorporated map guides into the first five rounds of competitive games, which will show you suggested angles for things like smoke grenades.

“We get it,” says Valve. “You could learn some smokes offline, but who has the time?”

It says these new “competitive map guides are more limited than the ones you use offline, but should be enough to give you a few reminders.” The feature is obviously optional and can be toggled in the pause menu, and launches with official map guides for the maps in competitive rotation: but it will also work with guides from the Steam Community Workshop.

Finally, friends can now join your custom game modes without any hassle, straight from the friends list. You simply start up a Workshop map, turn on the new “open party” option, and the game will be hosted on your PC and friends can join.

The map guides and friends stuff seem unequivocally good additions, but of course no-one’s talking about them. I’ll be curious to see how the new reloading mechanic feels tonight, and I suspect like many other CS2 players my muscle memory is going to get me killed an awful lot (I too am a chronic reloader, the kind that kills one player then gets taken out by his buddy because I’ve immediately spammed ‘R’).

But I welcome Valve’s willingness to tinker with CS2 in such a fundamental way: it might not work, and there might have to be a U-turn. But I can also see it adding something to the game’s many tactical layers, and if nothing else it’s going to force some of us to unlearn bad habits. Me: it’s going to force me to unlearn bad habits. The proof with this one really will be in the pudding.