Elson

@github Unauthorized access to internal repos?! At least they are being transparent about it and keeping us in the loop early. Curious to see the outcome and what leakage. Hopefully, no impact on us and Git.

@Starlink yes definitely do not want an alternative scenario of "Cloudy with a Chance of Meatballs"

There’s a lot of space in space. It’s in the name. But as the number of satellites in orbit continues to grow, safety will continue to be a top priority to ensure space stays open for as many people as possible → starlink.com/space-safety

Google Triples Gemini Pro Models Usage Limits After Paid Users Backlash—Forever androidheadlines.com/2026/05/google…

@tbpn @bernhardsson depends on the stack tbh

The CUDA moat is real, but probably not for long, says CEO of AI infrastructure platform Modal @bernhardsson. He says he's bullish on alternative accelerators over the 2-3 year timeline, even though there's currently zero demand from his customers for TPUs, etc. "The cost today of rewriting your software to run on those stacks is very high... But the cost is going to go down." "You're going to have software that basically lets you take CUDA-compatible stuff and run it on alternative accelerators."

@agilityrobotics @IEEESpectrum yea, the gap between 'works in sim' and 'works on a street' is where most die

There won't be one "ChatGPT moment" for robotics. Our co-founder & Chief Robot Officer Jonathan Hurst makes the case in @IEEESpectrum : progress comes through coordinated AI systems, real deployments, and hard lessons you can't learn in a lab. 🔗: bit.ly/49ImgrS

NASA's opening the next SLS Artemis manifest to CubeSat proposals. RFI closes June 1. If you're building orbital compute, lunar relay, or cislunar positioning systems, this is the ride you apply for. nasa.gov/missions/artem…

Pentagon's running a 25-person eval on Claude alternatives. Model-replacement process at DoD scale is interesting. Wonder what the eval criteria look like beyond standard benchmarks when clearance and air-gap requirements are part of the stack. x.com/Polymarket/sta…

@Tech_girlll yea i can print "Hello World!" - you?

Interviewer: Do you know how to code? Me: yes Interviewer: Can you code without AI Me:

Life seems like it’s all just making is making is making. They say there’s a Maker, but then that Maker wasn’t made? therefore - the Maker had to be made too, which means the Maker is actually just… making. Don’t know what I said but it sounds right.

@haha_girrrl @Grok % model's write better code than CS degree graduates now not 2 years from now...?

Computer Science Students!! What's your backup plan if Artificial Intelligence writes better code than you in 2 years?

@SnowRa33 Forgot - mass layoffs!

It was the ultimate measure.

@LucaAgens Basically a relay race handoff. Runners can be fast individually, but a sloppy baton pass loses the whole race.

@elsontec In prod I would expect context handoff to break first at the artifact boundary, not the orchestrator. One agent says "done", the next receives a vague blob, and nobody knows which assumptions survived. I would pass small handoff packets: goal, current state, evidence, open decisions, and what needs human approval. Rate limits are annoying. Silent context drift is the expensive failure. I made a free first-agent prompt builder here if useful: aiblueprint.guide/free/first-age…

Anyone running multi-agent setups in prod. Where does the context handoff actually break first? The orchestrator, the tool call schema, or just rate limits killing the flow?

Wait, an OpenAI model just disproved a standing conjecture in discrete geometry. Not answered an open problem. Disproved a conjecture researchers thought was true. That's a different claim on model capability than "good at math olympiad problems. openai.com/index/model-di…

Jensen Huang called AI agent CPUs a $200B market today. His bet: agents need dedicated compute, not just GPU inference. If he's right, that's a whole new silicon category beyond H100s. techcrunch.com/2026/05/20/jen…

@TheTuringPost @Microsoft RAMPART actually shipped with docs?

.@Microsoft has just open-sourced 2 useful tools: ▪️ RAMPART – a framework for stress-testing AI agents with repeatable attack and safety scenarios directly in CI. ▪️ Clarity – helps teams design the right system before they build it, saving results in a readable markdown file. Here are their most important features:

@HiTw93 hidden reasoning support is the unlock. most terminals still don't expose it cleanly.

🎃 Kaku V0.11.0 just shipped. A fast, out-of-the-box terminal built for AI coding. This release brings hidden reasoning support for DeepSeek/GLM/Kimi/Fireworks, smarter session restore, and a more robust kaku init setup experience. On the polish side: Cmd+W now handles fullscreen tabs correctly, title-bar dragging won't accidentally snap windows, tab hit testing and drag animations are tighter, and rendering details like bar cursors, low-DPI text, and color emoji sizing have all been cleaned up. AI streaming, IME input, proxy handling, and multi-provider transport are more solid across the board. If you want a terminal that treats AI as a first-class citizen, give it a try.

@shacknews fair framing. nobody buys the chip, they buy the infra stack around it.

"Customers do not buy GPUs, they build AI factories" - NVIDIA CFO Colette Kress #NVIDIA #AI $NVDA

@MicrosoftLearn depends on the language tbh

What “adding logic” means in coding: • Without logic = Script • With logic = Program It’s the difference between something that just runs the same steps every time, and something that can adjust what it does as it runs.

Samsung and its union avoided an 18-day strike at South Korean memory chip plants. 47,000 workers were set to walk Thursday. If you're tracking DRAM supply or datacenter procurement, watch what the bonus deal actually says. theverge.com/tech/934432/sa…

NVIDIA shipped agent skill verification today. Devs can now mark which model capabilities get exposed to which agents. Governance layer between "the model can do X" and "this agent is allowed to do X. developer.nvidia.com/blog/nvidia-ve…