PROJUKTI BRIEF
1 Meta's brain-to-text tool decodes sentences from raw brain signals at 61% accuracy.
The updated Brain2Qwerty v2 analyzes signals captured through a noninvasive scanner and decodes full sentences in real time. It is Meta's strongest brain-to-text model yet, trained on roughly 22,000 sentences from nine volunteers, each recorded for 10 hours. The stated goal is helping people who can no longer speak after a stroke, accident, or brain disorder regain a way to communicate. Learn more here.
2 OpenClaw and Cursor both launch mobile apps, putting always-on AI agents in your pocket.
OpenClaw, the platform that popularized personal AI assistants back in January, is now live on iOS and Android, letting you command your assistant from a phone-native app. Cursor followed with its own iOS app, so you can kick off always-on cloud coding agents from anywhere. This viral release video breaks down how it works (3M views).
3 Samsung, SK Hynix, and Micron face a class-action lawsuit over alleged price fixing.
A new California lawsuit accuses all three memory giants, who together control roughly 90% of the global DRAM market, of illegally coordinating to restrict supply and inflate prices. The suit lands as memory costs have soared over the past year, driving up prices on everything from Apple Macs to Microsoft's Xbox.
PROJUKTI FORWARD
The old calculus. For years, choosing between a closed model and an open-weight one was not really a decision. Proprietary models like Claude, ChatGPT, and Gemini were more powerful than open-weight alternatives by a wide enough margin that the choice made itself. That gap has closed abruptly over the past few months, and corporate America now has a real decision in front of it.
The difference. Open-weight models are published freely for anyone to download, which means you can host them privately, customize them, and keep your data on your own infrastructure. Proprietary models like GPT-5.5 and Claude stay closed: you access them only through the company that owns them, typically through a chatbot or an API.
Why teams are looking. Open-weight models offer more control, more privacy, and often lower long-term costs than proprietary alternatives. None of that mattered much when the models trailed badly on performance. That is no longer true. Zhipu AI's GLM-5.2 now ranks sixth on Artificial Analysis' leaderboard, closely followed by MiniMax-M3 and DeepSeek V4 Pro.
The budget pressure. Some companies are already feeling the cost of staying closed. Uber and ServiceNow blew through their entire annual AI budgets just months into 2026. Shifting select workloads to cheaper open-weight models that now match performance is one option teams are weighing to bring costs back under control.
Not a clean swap. The trade-offs in performance, security, and engineering overhead mean open-weight is not the right fit for every team or every workload. The real shift is not that everyone moves at once. It is that for the first time in years, the decision is genuinely worth making rather than skipping.
PROJUKTI SNAPSHOT
The better-alternative clause. It is frustrating when AI blindly follows your instructions even when there is a smarter option. One Redditor's fix: add a rule telling the AI to always suggest a better alternative before doing what you asked. Simple, and worth stealing (2,500 upvotes).
Five roles, not your job title. Claude Code's creator sent out a high-signal post on what tech roles could look like going forward. He expects five broad role archetypes to replace today's domain-specific job titles (2M views). Worth a read if you are thinking about where your role goes next.
The fake-profound tell. The internet has found a new giveaway for AI-written text: the fake profound statement. This thread breaks down the structure so you can spot it, and avoid it, if you are editing AI drafts before publishing them.
Tom and Jerry, live action. AI video tools have gotten good enough to convert 2D animation into convincing 3D live action. This viral Tom and Jerry clip shows exactly how far the conversion has come (3.5M views).
Keeping chatbots out of your business. National Design Studio released Rampart, an on-device tool that helps keep your personal information separate from AI chatbots. A practical option for anyone uneasy about what gets typed into a prompt box.
