@AshConnell Interesting take. If AI writes most of the code, the language that matters most might just be English, or whatever the developer actually thinks in.
@yacineMTB Commits used to measure output. Now the real work is in the prompts, the specs, and the reviews, and that barely shows up in a commit graph.
Nvidia just launched Nemotron 3.5, a new set of open models built for AI agents. The flagship Super VL has 120B parameters for enterprise agent work, plus two smaller Bolt models that run locally on AI PCs.
Nvidia just launched new open-weight models for AI agents: Nemotron 3.5 Super VL for enterprise use, plus two lightweight Bolt models that run locally on AI PCs. Available now, with Salesforce and EY among the early users.
@jbrooksuk The typing was always the least valuable part. Judgment about defaults, error states, and the human on the other end was the job. It's nice that this is finally visible.
@neerajjj6785 The real answer: the agent doesn't click either. It reads the page, makes the decision, and the booking API does the rest. The interviewer basically handed you your pitch.
@ClubofDreamers The 90-day lows might be the most useful part. Storage prices move quietly until they don't, and having the history in one place saves a lot of tab-hopping.
Microsoft says your workplace AI agents may be insider threats, even when no hacker is involved. It launched Guardian Agents to watch the AI helpers employees use, since overprivileged agents can act on their own.
Microsoft just pulled out of Nscale's UK AI data center in Loughton after grid delays pushed power availability into the 2030s. The 50MW site was set to run 23,040 Nvidia chips for Azure. Microsoft says the wider Nscale partnership continues.
Homeland Security just ordered every US civilian federal agency to strip DeepSeek products from its IT systems, including third-party tools built on DeepSeek models. Future procurement is banned too.
An open-source tool called REA lets AI coding agents like Codex and Claude Code take apart finished software and figure out how it works. It just blew past 327,000 GitHub stars, adding roughly 200,000 in a single week.