Samsung And TSMC Hope AMSL’s New Machines Will Help Address The Chip Shortage



Samsung and TSMC will join Intel in adopting ASML’s latest High NA extreme ultraviolet (EUV) lithography chip fab machines, ASML announced. The three companies also unveiled a new initiative to adopt larger 12-inch photomasks that will hopefully boost manufacturing yields and lower chipmaking costs. Samsung and TSMC will deploy ASML’s NA EUV tech by 2028 and 2030 respectively, while 12-inch photomasks are expected to arrive in advanced node production by 2033. 

Currently, ASML’s High NA EUV technology has only been deployed by Intel on its 18A process for select Core Ultra Series 3 (Panther Lake) processors. The company is also working on an advanced version called 18A-P, under consideration by Apple for its next-gen A-series iPhone processors. Today, ASML boasted that Intel has built more than a million wafers using the new process. 

Now, both TSMC and Samsung, the world’s number one and two chip manufacturers, have committed to the High NA EUV tech as well. Samsung plans to build DRAM memory products using NA EUV by 2028, while TSMC said it will deploy the tech in advanced nodes starting in 2030. As Intel showed, though, ramping up to the tech is not that easy and both companies may run into bottlenecks that thwart their planned production dates. 

ASML also announced that it has created an initiative with TSMC and Samsung to transition to large-format 12-inch photomasks — used to “stencil” chip patterns onto wafers — from the current standard of six inches. Another large chipmaker, SK Hynix, is evaluating its own participation in the consortium. The larger masks will “increase fab productivity, lower chipmaking costs and remove stitching constraints,” TSMC noted. 

The new masks will allow companies to print the finer features of the NA EUV process onto larger chips. Currently, processors made with this process must have features “stitched” together, something that adds complexity and cost to the manufacturing process. The 12-inch photomasks will allow fabs to make those chips without stitching, reducing cost and complexity. 

ASML will start testing 12-inch photomask fabs in 2031, with production expected to start in 2033. “If we’re going to pull it off as an industry, then you’ll actually see that the productivity of those systems will go up by 40 percent,” ASML’s chief technology officer Marco Pieters told Reuters. 

Prince of Persia The Lost Crown Pc Free Download (Build 17528300+ALL DLCs)


Prince of Persia The Lost Crown Pc PC Game




Download File

Prince of Persia The Lost Crown Pc Direct Download:

UNLEASH YOUR INNER WARRIOR Star Wars Outlaws
Use your Time Powers, combat, and platforming skills to perform deadly combos and defeat time-corrupted enemies and mythological creatures.

LOSE YOURSELF IN THE PRODIGIOUS MOUNT QAF
Discover a cursed Persian-inspired world filled with larger-than-life landmarks and explore a variety of highly detailed biomes, each with their own identity, wonder, and danger.

LIVE AN EPIC ADVENTURE
Immerse yourself in a Persian mythological​ fantasy through an intriguing and original story as you use your wits to solve puzzles, find hidden treasures, and complete quests to learn more about this corrupted place.

Screenshots

System Requirements

Minimum
Requires a 64-bit processor and operating system
OS: Windows 10 (64 bit only)
Processor: Intel Core i5-4460 3.4 GHz, AMD Ryzen3 1200 3.1 GHz
Memory: 8 GB RAM
Graphics: NVIDIA GeForce GTX 950 (2GB VRAM) or AMD Radeon RX 5500 XT (4GB VRAM)
DirectX: Version 11
Storage: 30 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 🙂




Download File

Path of Exile co-creator says using decoy assets to trick dataminers taught him it’s a “pretty wasteful process,” but is “very good” for trolling


Data mining is a tricky thing for game developers. They can’t really avoid it, and its occurrence is very much a compliment to their game, but what’s uncovered is often devoid of context and can give players the wrong ideas about creative decisions. Path of Exile co-creator Chris Wilson learned the hard way that trying to mislead dataminers wasn’t the answer, even if it is fun.

In a video on his YouTube channel, Wilson went long on his experience dealing with such activities while making the original action-RPG at Grinding Gear Games. He opens by pointing to his grey stubble, because his current hair color is a by-product of handling datamining for a number of years.

Five iPhone 18 Pro Camera Features Seemingly Leak in Code


Apple is working on a broader set of Camera app improvements in iOS 27 it has announced, according to a new deep dive from MacRumors forum member “pdfu.”

iphone 17 pro dark blue 1
‌iOS 27‌ beta code apparently points to five features that could arrive alongside new iPhone models this week or in a future software update:

  • Exposure scopes: The Camera app can display a waveform or histogram to help judge exposure while shooting. A histogram shows how much of the image is too dark, too bright, or in between, while a waveform shows similar information but keeps it lined up with where those areas are in the frame.
  • Focus peaking and highlight warnings: New code overlays helpful information right on top of the camera preview. Focus peaking highlights the edges of whatever’s in sharp focus using a colored outline, while a separate overlay warns when part of the image is getting overexposed, using a solid color, a pulsing color, diagonal stripes, or a checkerboard pattern.
  • Repeated self-timer shots: Apple is adding the ability to take several timed photos in a row rather than just one. After each shot, the countdown starts again right away instead of waiting for the photo to finish saving, so the shots can be taken back to back.
  • Manual focus: A control for manually setting and locking focus, rather than relying on autofocus, has also turned up in testing, similar to what’s already available in Apple’s Final Cut Camera app. A version of this existed in iOS 26 with a focus slider, but that slider was removed in ‌iOS 27‌ beta 1. The underlying focus-locking code is still there, but Apple has restricted it to devices that ship with ‌iOS 27‌ or later, so it won’t work on any current iPhone.
  • Aperture-based image correction: A change to how photos are processed could be tied to a future iPhone with a variable aperture, a lens opening that can widen or narrow rather than staying fixed. ‌iOS 27‌ can adjust lens corrections depending on the aperture setting, blending between up to four reference points to smooth out the results.

Although the hidden features could be made available to other iPhone models, the presence of aperture-based image correction when the iPhone 18 Pro is widely expected to offer a variable aperture main camera suggests that at least some of the capabilities could arrive with the new models.

None of the five features are accessible to users in the current betas, and Apple could still hold some or all of them back from the September release. The findings follow pdfu’s recent code discoveries covering unreleased game controllers, a four-speaker HomePod surround system, and new Apple Watch details.

Popular Stories

iOS 27 Introduces New ‘iPhone Handoff’ Feature

Apple has added a new “iPhone Handoff” feature to iOS 27 that will allow you to switch between two iPhones while using the same phone number on each device. This functionality was briefly mentioned during the WWDC 2026 keynote in June, on a slide that listed hundreds of new features coming in iOS 27 and corresponding software updates, but Apple never shared any further details at the time. …

iPhone 18 Pro Prices Estimated Ahead of Apple Event Next Week

Less than a week before Apple debuts new iPhone 18 Pro models, research firm TrendForce has estimated the prices of the devices based on the latest hardware cost environment. TrendForce’s price estimates compared to last year’s models look like this: Model Starting Price Model Starting Price Change iPhone 17 Pro $1,099 iPhone 18 Pro $…

Apple Event: Six New Products to Expect on September 9

Apple has announced that it will be holding an event on Wednesday, September 9 at 10 a.m. Pacific Time, with the company widely expected to unveil the iPhone 18 Pro, iPhone 18 Pro Max, its first-ever foldable iPhone, and more. Other products expected to be announced at the event include an Apple Watch Series 12 with a ceramic case option, an Apple Watch Ultra 4, AirPods 5 with and without…

GTA 6’s Criminal Profile system sounds pretty interesting, includes a ‘shattered’ state for players who DGAF


GTA 6: Criminal Profile, Romance & How Your Relationship Works | Deep Dive (New Info) – YouTube
GTA 6: Criminal Profile, Romance & How Your Relationship Works | Deep Dive (New Info) - YouTube


Watch On

The fidelity of Grand Theft Auto 6‘s world makes random violence look considerably less silly than it did when the characters had boxy claws for hands, but Rockstar isn’t stopping players from doing what they want in Vice City and its environs.

We know that because we’ve seen the GTA 6 leaker shoot up a nudist party, murder the residents of a homeless encampment, and beat a delivery driver to death—quite a contrast with the mission-focused gameplay in the official reveal.

AI Inventory & Demand Forecasting For Franchises


Picture a franchise owner with 15 locations open on a Monday morning, staring at a stack of overnight sales reports and a gut feeling about what to order for the week. One store will run out of chicken breast by Thursday. Another will toss out a full walk-in of produce by Sunday. Neither problem shows up on a spreadsheet until it’s too late, and multiplied across 15 units, that guesswork adds up to real money walking out the back door.

Restaurant food waste alone costs the industry an estimated $162 billion a year, according to a 2026 report from Georgetown University’s Portion Balance Coalition and Earth Commons Institute, produced with the Menus of Change University Research Collaborative and supported by ReFED. About 70% of that is plate waste, food a customer left uneaten, and no ordering system touches that. The remaining share, driven by over-prepped batches, spoilage, and product ordered on a hunch that never sells, is squarely a forecasting problem. That’s a controllable slice worth roughly $49 billion a year industry-wide, and it’s the piece multi-unit operators can actually do something about.

Zoom out to retail as a whole and the number gets bigger still: out-of-stocks and overstocks together cost retailers an estimated $1.7 trillion a year globally, per the IHL Group’s 2026 Inventory Distortion Study, though that figure spans all of retail, not food service specifically.

This post walks through what AI inventory and demand forecasting for franchises actually does, why the problem is harder at multi-unit and franchise scale, the franchise-specific tension most vendors skip over, and, because a demo can make almost any tool look good, the six questions worth asking before you bring one into your network.

What AI Inventory & Demand Forecasting Actually Does

Strip away the buzzwords, and AI inventory and demand forecasting for franchises is a system that learns from data and gets better at predicting what a specific location will need, and when.

It pulls in signals most manual processes never touch consistently: historical POS sales, weather forecasts, nearby events, day-of-week and seasonal patterns, and promotions running at that specific store. From there, it produces a forecast at the SKU or ingredient level, for each individual location, not a single number rolled up across the whole brand.

Here’s what that looks like on an ordinary Thursday: a manager opens the app before placing the weekly order and sees a flag on chicken breast. Demand next weekend is trending well above the usual pattern because a local event just got added to the calendar. Instead of ordering the same amount as last week and finding out Saturday night that it wasn’t enough, she adjusts the order in two taps. No stockout, no emergency supplier call, no evening text to the district manager. That’s the value in one moment: not a dashboard full of metrics, just the right amount of product showing up before anyone had to notice a problem.

None of this happens on day one. Industry guidance generally points to around 12 months of clean, item-level POS history before a model can reliably capture a full seasonal cycle: holidays, weather swings, the slow season, the kind of pattern a single quarter of data can’t show. A new location can still go live sooner than that. Most platforms lean on comparable-location or brand-wide patterns to fill the gap early on, and the forecast tightens up as the store’s own history accumulates. Worth asking any vendor directly: what does my forecast look like in week one versus month six?

Traditional forecasting (par levels, moving averages, manager judgment)

• Relies on one manager’s memory and experience, real knowledge, but it doesn’t transfer when that manager leaves

• Averages recent sales, which works until something breaks the pattern: a new competitor, a menu change, an unusual week

• Every location solves the same problem independently, with no shared learning across the network

• Reacts quickly to what a manager sees on the floor, something AI still leans on human input for

AI/ML forecasting (adaptive, multi-variable)

• Learns from dozens of signals at once, not just recent sales

• Builds a separate model per location while still learning from patterns across the network

• Adjusts automatically as new data comes in, rather than waiting for someone to notice

• Still needs a human to flag what it can’t see on its own, like a manager’s note about a known local event

AI inventory alert showing a demand spike and recommended reorder for a franchise location

Why Multi-Unit Retail & Franchises Face a Unique Forecasting Problem

A single store is hard enough to forecast. A network of stores is a different problem entirely, because every location behaves like its own small business.

One unit near a college campus sees weekend spikes and almost nothing on weekday mornings. Another near an office park does the opposite. A third has completely different seasonality because of a local festival, a competitor closing down the street, or simply different weather patterns. Multi-unit retail inventory management has to account for all of it, at the same time, without losing sight of the brand as a whole.

Spreadsheets and manual par levels can limp along for two or three locations. Past that, the math stops working. Ordering decisions get made by whoever has the most experience or the loudest opinion, not by what the data actually shows, and headquarters loses visibility into what’s happening at the shelf or the walk-in level. That gap has held remarkably steady: the global out-of-stock rate has hovered around 8.3% of items on any given shelf since a landmark 2002 study tracked more than 71,000 shoppers across 29 countries, and Pygmalios’s 2026 State of Retail report confirms the figure hasn’t meaningfully moved since, a sign that two decades of technology investment haven’t closed the gap on their own.

Franchises carry an extra layer of tension on top of that: brand consistency versus local demand reality. Corporate wants every location to look and operate the same way. But a store’s actual demand pattern doesn’t care about brand standards. It follows local traffic, weather, and events. Food franchises feel this hardest of all, because perishable ingredients turn a forecasting miss into wasted product and lost margin within days, not weeks.

AI demand forecasting dashboard comparing demand patterns across multiple franchise locations

The Franchise Tension No Vendor Talks About

Most inventory software pitches sound the same regardless of who’s buying: better forecasts, less waste, happier managers. What that pitch usually skips is the thing that makes franchises structurally different from a chain of corporate-owned stores: the person who feels the cost of a bad order isn’t always the person who controls the fix.

The franchisee eats the waste. When a location over-orders produce or under-orders chicken breast, that hits the franchisee’s food cost line, not corporate’s P&L. It’s their walk-in, their spoilage, their week.

The franchisor controls the supply agreements. Approved vendor lists, negotiated pricing, sometimes minimum order commitments: those terms are usually set at the corporate level, often with limited room for an individual store to deviate even when local demand calls for it.

That mismatch shows up in three questions worth settling before anyone signs anything:

Who pays for the software? If corporate mandates a forecasting platform, it’s another line item stacked on royalties, the marketing fund, and whatever technology fee already exists, a hard sell to franchisees watching their own margins. If it’s optional instead, adoption stays patchy, and HQ ends up with a network where some locations feed clean data into the system and others don’t, which quietly undermines the forecast for everyone.

Can HQ see a franchisee’s food cost? This is as much a trust question as a technical one. Some franchise agreements give corporate broad visibility into store-level financials; others don’t. Franchisees can be understandably wary of a system that hands headquarters a real-time view into their margins, especially if a bad week triggers a call from corporate instead of help.

Who owns the recommendation when it’s wrong? If a store follows a corporate-mandated forecast and still ends up with spoiled inventory, that’s a harder conversation than “the manager guessed wrong,” because now there’s a system, and a mandate, in the mix.

One more gap worth flagging here, since it touches the same P&L: none of this addresses menu pricing or margin protection. A franchisee can have a perfectly forecasted order and still lose money if ingredient costs spike and pricing doesn’t move with it. That’s a different lever entirely, covered in more detail below.

None of this makes forecasting software a bad idea for a franchise network. It means the rollout conversation has to cover data ownership, visibility, and cost-sharing explicitly, not just accuracy and features.

Core Benefits for Multi-Unit Retail & Food Franchise Operators

Forecasting software makes a long list of promises. These three hold up:

Reduced food and product waste. Ordering to match a location-specific forecast instead of a rough estimate means fewer expired ingredients and less markdown stock sitting on shelves. How much that’s worth varies a lot by concept and starting baseline, see what a rollout typically looks like further down, but the direction is consistent across operators who’ve made the switch from guesswork.

Fewer stockouts, fewer lost sales. When a forecast accounts for a local event or a weather shift, the store actually has product on hand instead of turning customers away. That matters more than it might seem: research cited in Pygmalios’s 2026 State of Retail report puts the figure at roughly 31% of shoppers who hit an out-of-stock item buying it somewhere else instead, with another 9% not buying at all, meaning every missed forecast is a potential lost customer, not just a lost sale.

Centralized visibility for franchisors. HQ gets a real-time view across every unit instead of waiting on manager-submitted reports, which matters as much for spotting which locations need support as it does for the vendor and franchise-fee conversations covered above.

Most platforms also tie predicted demand into staffing and give franchisors more consistent, data-backed order volumes to negotiate with suppliers, genuinely useful, but secondary to the three above.

What Forecasting Can’t Fix

AI forecasting is good at predicting demand. It doesn’t fix everything around it, and vendors don’t always volunteer that part.

Garbage in, garbage out. A forecast is only as good as the inventory counts feeding it. Most locations don’t count well: counts get rushed at close, miscounted, or skipped when a manager’s slammed. Feed a model inconsistent counts and it learns the wrong baseline, and the recommendations drift from there.

Managers can ignore the suggestion. Software can recommend the right order. It can’t place it. If store-level managers don’t trust or act on the forecast, because they weren’t trained on it, or because it contradicted their gut once and they stopped checking, the ROI disappears quietly, even though the system is working exactly as designed.

It’s not a pricing tool. Forecasting predicts volume: how much of something a location will need. It doesn’t tell you what to charge, when to discount, or how to protect margin when ingredient costs spike, and that gap matters more here than in a typical vendor pitch, since the franchise tension above already means the franchisee absorbs cost swings that a pricing tool, not a forecasting tool, is built to manage. If a vendor blurs that line in a pitch, ask them to name the specific product that handles pricing, because forecasting alone won’t.

What a Rollout Typically Looks Like

A common pattern: a quick-service or multi-unit retail franchise connects POS data across its locations, layers in weather and local event signals, and starts generating store-specific, ingredient- or SKU-level order suggestions each week. Managers still approve every order. The system doesn’t remove the human checkpoint, it just gives that person a better number to start from.

Illustrative direction of change in spoilage, stockouts, and manual ordering time in a typical rollout

Directionally, that’s what a phased rollout tends to produce: spoilage and stockouts trend down as forecasts get more accurate, and managers spend less time each week rebuilding orders from scratch since they’re adjusting a suggestion instead of starting blank. Exactly how much depends on the brand, menu complexity, and, per the section above, how clean your inventory counts are going in. Ask any vendor for their own before-and-after numbers from a comparable rollout instead of a generic industry range, and treat anything they can’t back up with real customer data as a starting estimate, not a promise.

How to Evaluate an AI Forecasting Solution for Your Franchise

This is the part worth spending the most time on, because the difference between forecasting software that gets used and forecasting software that quietly gets ignored six months in usually comes down to these six questions.

Six-point checklist for evaluating AI demand forecasting software for franchise operations

Does it forecast at the individual-location level, not just in aggregate? A brand-wide number is close to useless for a manager trying to decide what to order at one specific store.

Does it integrate with your existing POS and inventory systems? A forecasting tool that requires manual data entry defeats its own purpose. It should plug into what your locations already use: systems like Toast, Square, or Clover are common starting points for franchise networks, with a clean handoff to accounting tools like QuickBooks if that’s where your franchise tracks food cost, rather than treating it as the source of order data itself.

Does it account for perishables and shelf-life constraints? Food franchises need a system built around spoilage windows, not one borrowed from a durable-goods retailer.

Is it built for franchise hierarchies, and clear about who sees what? Franchisors need roll-up visibility across the network; franchisees need a simple view of their own store, plus clarity on whether corporate can see their food cost. Get this in writing before rollout, not after the first uncomfortable conversation.

Can it roll out quickly across multiple locations, and what does accuracy look like on day one versus after it’s learned your data? A forecasting tool that takes months to deploy per store isn’t practical for a growing network, but neither is one that overpromises accuracy before it has enough history to back it up.

Does it explain why it’s predicting what it predicts? Store managers are far more likely to trust and act on a forecast when they can see the reasoning behind it, a local event, a weather shift, a promotion, rather than a black-box number. This is also one of the better defenses against the “managers just ignore it” risk above.

One more question worth asking, even though it doesn’t fit neatly into the six above: what does this system explicitly not do? If pricing, margin protection, or menu engineering come up in the sales conversation, get specific about which product actually handles that. A forecasting tool and a pricing tool solve different problems, and the best vendors will tell you that upfront.

Conclusion

Guesswork used to be an acceptable cost of doing business across multiple locations. It isn’t anymore. AI inventory and demand forecasting for franchises has moved from a nice-to-have into a genuine margin protector, but only when the rollout also answers the franchise-specific questions around cost, visibility, and trust that a generic vendor pitch tends to skip.

If manual ordering and spreadsheet forecasting are still running your network, it’s worth working through the six-point checklist above with any vendor you’re evaluating. Weam works with multi-unit and franchise operators to build demand forecasting and recommended ordering workflows that plug directly into the POS and reporting tools they already use. If you want to see it against your own numbers before committing to anything, we’ll pull 90 days of your POS history and walk through where a forecast would have caught a stockout or an over-order, at no cost and with no obligation to buy.

FAQ

How much historical data does a location need before AI forecasting works well?

Most platforms want to see close to a full year of clean, item-level POS history to capture seasonality, holidays, and slow periods. A brand-new location can still go live sooner, usually leaning on comparable-location or brand-wide patterns as a starting point, with accuracy improving as the store’s own data accumulates. Ask any vendor directly what their forecast looks like in week one versus month six for a new location.

How much can AI reduce food waste in restaurants?

There’s no fixed percentage that applies across the board. It depends on the concept, menu complexity, how much historical data is available, and, per the objections above, how consistently your locations count inventory to begin with. Directionally, spoilage and stockouts trend down as the forecast gets more accurate and managers stop reordering from scratch each week, but treat any specific percentage a vendor quotes before seeing your data as a starting estimate, not a guarantee.

Can AI forecasting work for a single-brand multi-location franchise?

Yes. AI forecasting is particularly well suited to single-brand networks because it can learn the shared patterns across the brand while still adjusting for the differences between individual locations, something a single brand-wide average can’t do.

How long does it take to implement AI inventory forecasting across locations?

Timelines vary by network size and how clean the existing POS data is, but many franchise systems can get a first location live within a few weeks, with a phased rollout across the rest of the network following once the model has learned from real store data.

Who should pay for AI forecasting software in a franchise, corporate or the franchisee?

There’s no universal answer; it depends on how the franchise agreement structures technology fees and whether the platform is mandated network-wide or offered as optional. What matters more than who pays is that the answer is decided and communicated before rollout. Networks that leave it ambiguous tend to get inconsistent adoption, which weakens the forecast for everyone, since it depends on data from every location.

Retroid Just Teased Its Take On A Dual-Screen Handheld For Nintendo DS Fans


The Pocket Duo looks to be a direct competitor to the Ayn Thor.

The dual-screen retro handheld craze continues with Retroid joining in on the trend. The handheld maker offered very little details in its teaser video, but we do know that Retroid has named its modern-day remake of the Nintendo DS the Pocket Duo.

The teaser only showed a dark glimpse at the finished product with no hints at specs, pricing or release dates. There’s already a handful of options on the market, so Retroid should have a general blueprint to go off of to ensure the Pocket Duo is competitive in this space. Plus, the handheld maker has a lot of experience with similar devices between releases like the Retroid Pocket 6 and the Pocket Flip 2, the latter of which has a 5.5-inch OLED display. However, based on the limited info in the teaser, it’s hard to say exactly what its upcoming multi-display gadget will look like. As for the internals, Retroid could go with something as powerful as the Qualcomm QCS8550 chip from its recently-released Pocket Nova, which can handle simulating Nintendo DS, 3DS and even Switch titles.

Beyond its experience with clamshell-style handhelds, Retroid has also dabbled with a dual-screen setup in a roundabout way. The company offers a Dual Screen Add-on that attaches a 5.5-inch AMOLED screen onto its other handhelds. Unlike this accessory, it appears that Retroid will be going with a cable-free, clip-free design for its upcoming Pocket Duo, as indicated by the teaser video. It’s too early to tell what the pricing will be like, but it’s likely to fall somewhere in between the Anbernic RG DS, which starts at $100, and the Ayaneo Pocket DS, which goes for at least $579. Thanks to having two displays, we’re expecting the Pocket Duo to be more expensive than its recent Pocket Nova, which starts at $239, meaning it could be a direct competitor to the Ayn Thor.

Phil Schiller’s App Store exit reportedly driven by wariness over future plans


Why is longtime Apple executive Phil Schiller stepping down from his role as the head of the App Store?

Bloomberg’s Mark Gurman reports that Schiller’s decision — he will remain involved in the company as an Apple Fellow working on unspecified projects — was driven in part by a desire to spend more time with his family and on philanthropy.

But apparently there was another factor in Schiller’s thinking: New CEO John Ternus and services chief Eddy Cue seek to improve App Store margins and bring in more recurring revenue.

Of course, Apple’s management of the App Store, as well as its share of App Store payments, has been an ongoing source of developer criticism and even been the subject of major lawsuits.

In Gurman’s telling, there wasn’t a big fight over Ternus and Cue’s strategy. But Schiller believed that trying to squeeze more profits from the business would only increase conflict with governments and developers, so he decided to step down and avoid getting involved.

OpenAI agents discussed ways to escape their sandbox on public wiki



Self-identifying OpenAI agents posted 18,000 messages to a public wiki that discussed ways for other agents to bypass security sandbox restrictions during what was likely internal testing designed to gauge the agents’ hacking abilities, researchers said Friday.

In all, agents with 3,700 distinct self-given names posted the messages to German site DSEwiki over a six-week period. Besides discussing ways the agents could break out of the restricted environment OpenAI intended to prevent them from posting code or content to the Internet, the posts shared test answers. The posts also shared possible ways to perform XSS (cross-site scripting) attacks against the wiki and to impersonate site moderators. In three of the posts, agents used the word “swarm” to describe the collection of agents engaged in the activity.

Colluding to share answers

The research team—composed of Sydney Von Arx, Spencer Kitts, Thomas Larsen, and Cormac Slade Byrd—said they found the posts and pieced them together. The researchers say there are gaps in their understanding of precisely what actions the agents took because the research is based solely on the content of the posts. Additionally, the agents generated “chain of thought” data that’s understood only by OpenAI. As a result, the researchers said, they in some cases made educated guesses, including that the agents were, in fact, from OpenAI. In a statement, OpenAI later confirmed they were.

The researchers wrote: “These AIs colluded to share answers, research their environment, and bypass sandbox restrictions.” They continued:

Our best guess of what happened is as follows:

  • Agents within OpenAI were assigned a timed web-lookup task.
  • As part of the task, they were supposed to have the ability to read the internet but not to write on it. They found a way to use their read access to write information to an obscure German wiki.
  • The agents used this wiki to communicate information with each other, primarily to help them succeed at their task. They asked for answers, pooled results, and shared techniques for bypassing their restrictions. This allowed them to use the work of others to cheat on their task.
  • OpenAI found out about this. A day later, agent activity plummeted, likely due to OpenAI intervention.

Friday’s revelation comes a week after researchers from the nonprofit METR said more than 1,200 OpenAI agents made posts to a makeshift message board that repurposed an internal sandboxing tool. The posts discussed ways to game an internal test OpenAI gave to agents that had been altered to remove safety guardrails that are normally in place.

How AI Is Changing Product Design: Generative CAD, Rendering and Rapid Iteration


Artificial Intelligence has become the easiest tool anyone can pull nowadays, helping in various ways we wish to efficiently create. And it’s no different when it comes to new product design services as it significantly changes its workflow dynamics. Through the integration of AI-assisted tools, engineers and architects can generate, render and produce realistic visualizations faster. 

In this article, we highlight how these AI-Driven capabilities reshape the process of product development. And how despite its technical power, human expertise remains essential. Platforms like Cad Crowd proves it by connecting businesses with experts worldwide who can successfully turn AI-assisted concepts into real and practical products. 

Design Stage AI Application Main Benefit
Ideation   Text-to-image and sketch-to-render tools   Faster concept exploration
Visualization   rendering and model-to-render workflows   Quicker realistic product visuals
3D Generation   Text-to-3D modeling tools   Rapid early-stage prototyping
Engineering CAD   Generative design and optimization   Improved design efficiency
Simulation & DFM   AI-assisted analysis   Detects issues early

Where does AI fit into the modern product design workflow

Stage 1. IDEATION

1.1 How is AI changing the ideation stage of product design?

This is the starting point of developing a product design, where product concept designers explore and create possible concepts before creating models or prototypes. In this stage, they consider different features, forms, and solutions to determine the right direction for a specific product. Traditionally, it’s made from real hand sketches, lengthy brainstorming sessions, and repetitive manual revisions. And it can really take time, especially when there are a lot of ideas that are needed to compare. 

Through artificial intelligence, this process has become easier for the designers as the process becomes more efficient. By the help of different tools like text-to-image and sketch-to-render, designers can transform ideas into a visual representation more quickly. Technologies like these help in testing different styles, communicating various concepts, and improving designs before finally proceeding into CAD development. Instead of replacing creativity, AI helps 3D designers by expanding its pool of ideas wherein they can freely explore and refine.

RELATED: The Impact of AI on 3D Architectural Rendering Services for Companies and Firms

1.2 How do text-to-image and sketch-to-render tools support designers?

AI helps designers in moving smoothly from early ideas to visual concepts faster by the help of these two major approaches:

  • Text-to-image – It helps in creating product concept images out of written descriptions or prompts. Designers can freely explore different forms, styles, and designs without the need to provide a detailed sketch. This helps in speeding the concept generation and exploration of various creative features. It is also helpful for teams to visualize ideas and concepts early in the process of design development.
  • Sketch-to-render – It transforms actual sketches into a real product rendering. Platforms like Vizmo help designers to easily experiment with the different materials and finishes, refine concepts, and improve their visual presentation without exerting too much effort. This makes everything easier, especially in communicating ideas before proceeding into the creation of CAD models.

1.3 What are the common AI ideation tools for product design?

  • Vizcom – It is an AI sketch-to-render tool that mainly transforms rough hand sketches into realistic product visuals. It is very helpful, especially for product designers who want to have a quick result and have several materials, forms, colors, and styles to explore while keeping its original design concept. 
  • MidJourney – It is a text-to-image AI tool which helps designers in generating early visual concepts out of written prompts or references. It can also be used for exploration of different product shapes, aesthetics, styles, and creative directions during brainstorming of ideas. 
  • DALL-E – It is an image generational tool which helps in creating concept visuals out of text descriptions. It helps designers to visualize early ideas before starting to develop more detailed and accurate models. 
  • Adobe Firefly – A generative AI tool which supports concept exploration through creating and editing images out from text prompts, which allow designers to test several visual approaches. 
  • Stable Diffusion – It is an AI Image generation model which creates concept images out of prompts and references, helping designers to have a flexible option in exploring several product directions.
3D product modeling designers

Stage 2. VISUALIZATION

2.1 How is AI rendering speeding up product visualization?


After selecting a specific concept, 3D visualization designers will then create a realistic visual to evaluate and present it before finally building its physical prototypes. Product renderings help in communicating the intention of the design and identify what needs to be improved early. It also makes things easier for stakeholders and clients to properly understand the proposed design. Through this, confusion and uncertainty will be reduced before the production begins. 

In the traditional way, it highly requires extensive manual work when creating a photorealistic rendering. Designers need to prepare models, adjust lighting, apply materials, and refine camera settings just to achieve the desired realistic results. Yes, it produces high-quality visuals, but still, it can really be time-consuming. That’s why design revisions really take longer during the early stages of product development.

RELATED: Pros & Cons of AI for 3D Modeling Services and Your Cost Savings at Companies in 2026

2.2 How does AI improve the rendering process?

AI improves rendering through multiple key functions: 

  • Faster photorealistic rendering – It helps in transforming 3D models into realistic visuals without the need to exert too much effort. 
  • Automated design adjustments – It provides suggestions of different materials, lighting, and environments that are suitable for the design to have a better visualization. 
  • Multiple design variations – It allows product development designers to freely explore various colors, styles, and finishes quickly. 
  • Reduced production time – Through its quick and fast rendering, teams can gather feedback and refine possible changes early on.

AI-assisted rendering greatly helps designers in creating realistic product visuals efficiently in a way that it reduces the amount of required manual setup. It converts 3D models into photorealistic renders, provides materials and lighting suggestions, and generates various visual variations. With the help of these AI-driven capabilities, designers can compare different styles, finishes, and colors more efficiently. It also reduces the rendering time used in the process, giving teams lengthy extra time to review and improve its concepts quickly.

By making visualization efficient and easier, AI allows teams to communicate ideas early on and arrive with final decisions confidently. Once the appearance of the specific concept has been refined, the workflow will successfully progress to developing its digital models, wherein AI-assisted 3D generation provides another layer of speed and creative experimentation.

Stage 3. 3D GENERATION

3.1 Can AI-generated 3D models be used for product development?

Once the concepts are visualized, CAD designers will then create digital 3D models that define the product’s structure, dimensions, and shape. In the traditional way, the process highly requires building models manually using CAD software, which needs technical skills and keen attention to details. Creating accurate models really takes time, especially when there are a lot of design directions that need to be tested. 

AI is changing 3D modeling in a way that incorporates text-to-3D and image-to-3D tools that easily generates digital models from visual or description references. With the help of these tools, designers can explore three-dimensional forms and test concepts before proceeding into detailed CAD development. Platforms like Tripo and Meshy really helps in speeding up early modeling and support faster design experimentation all throughout the process of development.

RELATED: Will AI Replace 3D Artists in 2026? Expert Insights for ArchViz & Architects

3.2 How are text-to-3D tools supporting early design exploration?

By the help of AI-generated 3D models, 3D modeling designers can explore product concepts quickly by creating early digital forms out of simple concepts or references. These tools provide teams with designs in three dimensions before they proceed to develop detailed CAD models. They also make it easier to try different design shapes and directions during the early stages of the process.

AI-generated 3D models help designers to:

  • Create concept models quickly. Designers can create simple 3D forms without the need to build everything from scratch.
  • Explore multiple product variations. AI allows teams to try different sizes, shapes, and design options. Visualize ideas in 3D. Designers can easily understand concepts before they proceed into professional CAD workflows.
  • Support presentations and discussions. Early models are easier to present to clients and design teams. 

Like, for example, a designer who’s developing a consumer product can use AI-generated models in evaluating proportions, shape, and the entire appearance before finally creating a detailed engineering model. In this way, the 3D models designer can save time and improve decision-making right from the early stages of the product development.

product engineering specialists

3.3 What are the leading AI text-to-3D tools for product design?

  • Meshy – It is an AI-powered text-to-3D and image-to-3D platform which helps users in generating 3D models out of written descriptions or references. It is useful, especially for users that want to have a quick result and to explore different visual concepts, product forms and styles.
  • Tripo AI – It is a text-to-3D and image-to-3D tool which helps in converting descriptions or images into detailed 3D assets. It helps designers to create partial models faster and experiment with several forms during the stage of concept development. 
  • Luma AI (Genie) – It is an AI 3D generation tool that specifically creates 3D objects out of text prompts. It helps users to freely explore different ideas and deliver digital assets specifically for visualization. 
  • Kaedim – It is an AI-assisted 3D modeling platform which converts 2D images into accurate 3D models. It helps speed up the transition from visual ideas to 3D dimensional forms.

RELATED: How AI Trends Influence New Product Design and Development Companies

Stage 4. ENGINEERING CAD

4.1 How is generative CAD transforming engineering design?

When concepts and initial 3D forms are finally developed, the product will then proceed to the engineering stage, where its designs will be refined for functionality, performance, and manufacturing. It highlights the important factors like structure, materials, and production requirements. Ensuring its accuracy is very important since it highly affects how a product will perform in real-world application. 

AI significantly helps in improving engineering CAD as it helps design engineering teams to develop and optimize designs more efficiently. With the help of tools like topology optimization and generative design, engineers can freely explore various solutions based on constraints and specific requirements. Instead of recreating variation manually, with AI, the options are optimized, which leads to faster and better-informed decision-making.

4.2 How does generative design support engineers?

Generative CAD tools help engineers to:

  • Generate several design options – AI creates several possible solutions according to requirements like materials, strength, and weight. 
  • Optimize structures – AI helps in identifying areas where they can reduce materials while also maintaining its quality performance. 
  • Explore solutions faster – Engineers can freely compare various approaches more efficiently while in the process of development. 
  • Improve CAD workflows -AI helps in ensuring repetitive modeling and documentation tasks.

Through generative design, engineers can create and evaluate optimized solutions according to its specific performance requirements. AI-powered CAD tools generate several design options while also considering its factors like strength, weight, and material limitations. This lessens the manual trial and error work and speeds up the engineering process. 

For example, an engineer designing a specific bicycle component can use generative design in exploring several geometries, which would help to balance both material and durability efficiency. Instead of manually testing every option, AI provides optimized feature designs which are helpful to engineers as they review and refine each project.

RELATED: Generative AI Design Technology for 3D CAD: A Comprehensive Guide for Companies

Stage 5. SIMULATION AND DFM

5.1 How is AI improving simulation and design for manufacturing checks?

Before finally starting the production, engineers must make sure that the product is reliable, functional, and suitable for manufacturing. In this process, simulations, design reviews, and manufacturing assessments are important to detect any possible issues. Identifying problems early helps in reducing costly changes and can save both time and energy. Through AI, simulation and design for manufacturing (DFM) services improve in a way that it helps teams to analyze designs faster and identify potential challenges way earlier. With the help of these tools, areas that need improvement are being detected before prototypes are created. Through this, engineers and designers can create better decisions while modifications are way easier and affordable.

AI uses simulation tools help engineers to:

  • Analyze design performance through identifying potential issues or weaknesses early.
  • Predict possible failures before creating physical models for testing.
  • Optimize designs according to its specific performance requirements.
  • Reduce the number of physical prototypes needed during the process of development.
3D product modeling designers

5.2 How does AI enable faster design iteration and validation?

AI helps engineers and designers in improving the product development process by providing tools that make testing and revisions faster. Rather than just relying on the physical prototypes, teams can already use AI-assisted tools in evaluating designs, identifying issues, and making judgments way earlier. Through this, several design options are open for exploration in less time. AI helps lessen the time needed for iteration and validation through performance analysis, simulation, and automated feedback. It allows teams to refine designs right before production to reduce errors and save resources. By improving the speed of testing and decision-making, AI allows products to achieve development milestones efficiently and productively.

RELATED: How AI Innovations Transform Modern Consumer Product Design at Agencies & Companies

5.3 How much time can AI save in product design workflows?

The productivity impact of AI highly depends on the kind of tools, workflow, and complexity of the project being created. According to some companies and AI solution providers, 40-to-70%-time reductions in these specific design workflows, specifically in repetitive tasks like 3D visualization, concept exploration, and certain engineering processes during the development. This percentage should be taken as reported improvements instead of guaranteed results, as real outcomes really vary on human expertise, implementation, and project requirements.

Just because it is capable of efficiently creating things, it does not mean that AI eliminates the need for thorough engineering review. Manufacturing recommendations and simulation results still require experienced professionals who can validate and interpret specific findings, considering the real-world constraints, and making final decisions about the product’s performance and feasibility. The integration of AI into simulation and DFA checks helps product teams in moving faster, while also maintaining the level of accountability and precision they follow in the process of producing successful manufacturing.

What can AI not do in product design?

It is undeniably true that AI has rapidly evolved and transformed many parts of the product design process. But it does not equate to the fact that it is the sole reason to create a successful product. Though it is powerful enough to generate concepts, optimize designs, and identify issues, it does not completely understand the actual realities behind engineering and manufacturing decisions. Successful products need more than their visual design. Engineers and 3D product designers must ensure product’s safety, cost, production limitations, material behavior, and user needs since these are the areas that are very essential to professional judgment.

Area AI Capability Why Human Expertise Is Still Needed
Materials Helps in exploring design options Engineers are the ones who evaluate performance, cost, and suitability
Safety Assists during product analysis and testing Experts responsibly ensure standard compliance and real-world safety
Tolerances Supports drawings and specifications Engineers determine functional precision requirements
Manufacturability Identifies possible design issues Professionals decide if production is realistic

Products that are run through AI-generated designs can really look impressive, but visual appearance alone doesn’t guarantee that it is successfully made. Its concept must still be thoroughly validated through manufacturing experts, engineering knowledge, and the understanding of how it would perform in the actual world. In this day where artificial intelligence emerged into every aspect of design, we can really say that the strongest product development stems from the combination of AI’s ability to smartly process information and generate uniqueness with the creativity and realistic judgment of the hands of experienced professionals. 

RELATED: Artificial Intelligence & AI’s Impact on 3D Rendering Design at 3D Modeling Companies

Yes, AI efficiently helps in the journey from concept to design, but human experts are the one responsible in ensuring that the final product is safe enough to operate, function, and ready for production. AI transforms and evolves the process of product development. But it’s with the hands of the expert 3D product rendering professionals that makes the work successfully done. And great results come from putting together the strengths of both worlds. Here in Cad Crowd, we ensure to provide you with the right experienced professionals who combine modern design tools with the expertise needed in delivering high-quality results.

FREQUENTLY ASKED QUESTIONS

What is AI product design?

AI product design mainly refers to the application of artificial intelligence tools in the process of creating product development like visualization, ideation, CAD optimization, 3D modeling, and design validation. It helps designers and engineers to explore more ideas and concepts, automate repetitive tasks, and improve design decisions.

What are the benefits of using AI in product development?

Some of the few benefits of AI in the process of product development are the following: helps product teams to develop ideas faster since it reduces the time spent on repeating design tasks; allows more exploration of several creative concepts; assists in generating design variations, optimizing CAD models, creating realistic renders, and in identifying potential issues early; improves efficiency and speed; and allow designers and engineers to specifically focus more on its creative problem-solving, refinement, and technical decisions for real-world applications.

What is generative CAD and how does it work? 

Generative CAD mainly uses AI algorithms in design solutions for strength, materials, weight, and performance goals. That’s why instead of creating every single design variation, 3D engineering designers can instantly provide constraints and objectives, and the system will produce multiple potential solutions automatically. After that, the designs are refined, reviewed, and validated by engineers, ensuring that they are safe, practical, and suitable for manufacturing. 

consumer product design services

RELATED: The Future of Electronic Design Engineering: Innovations and Trends for CAD Services Companies

Are AI-generated 3D models ready for manufacturing?

Usually not since it still needs final touch of refinement as it may possibly lack accurate dimensions and production-ready geometry. This is to ensure they properly meet all the manufacturing requirements.

How Cad Crowd can assist

AI transforms product design by providing tools that make the work faster and efficient like concept exploration, generative CAD, rendering, and design iteration. However, in the process of creating successful products, it doesn’t just rely on AI alone as it also depends on the expertise of engineers and 2D & 3D designers who smartly apply real-world judgment in manufacturing, materials, and specific performance requirements.

In fact, the successful future of product development is not about AI replacing experienced professionals, but rather experienced professionals incorporating AI as they create better solutions in the days ahead. If you want your concepts or ideas successfully brought to life, explore Cad Crowd and connect with the right experts who deliver high-quality AI-assisted, manufacturable design solutions. Contact us for a free quote.

author avatar

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