today's signal · no scroll
the brief
Agents and coding tools drove the day: Microsoft unveiled big MAI moves in reasoning and code, Hugging Face highlighted fast local computer-use agents, and Claude Code shipped meaningful agent reliability and telemetry updates. Open-source saw a strong showing with Pluto.jl 1.0 and a new coding-agent interface, while a Stanford Law study and a clever VRAM swap hack rounded out a busy pulse.
An experimental nbd-vram driver turns spare GPU VRAM into swap space, offering a last-resort buffer against OOMs at the cost of risk and instability—handle with care.
Use your Nvidia GPU's VRAM as swap space on Linux — Article URL: https://github.com/c0dejedi/nbd-vram Comments URL: https://news.ycombinator.com/item?id=48377404 Points: 151 # Comments: 44
Microsoft introduced MAI-Thinking-1 (1T parameters, 35B active) and MAI-Code-1-Flash (137B, 5B active) among a new MAI family, with early access for select partners.
Microsoft's new MAI models — <p>Microsoft <a href="https://microsoft.ai/news/building-a-hillclimbing-machine-launching-seven-new-mai-models/">announced two new text LLMs</a> this morning - <strong><a href="https://microsoft.ai/news/introducing-mai-thinking-1/">MAI-Thinking-1</a></strong> (reasoning, 1T parameters, 35B active, available to "select early partners") and <strong><a href="https://microsoft.ai/news/introducingmai-code-1-flash/">MAI-Code-1-Flash</a></strong> (137B Parameters, 5B act...
Holo 3.1 brings much faster, local-first computer-use agents that can operate apps on your machine with lower latency and improved privacy for day-to-day workflows.
Holo3.1: Fast & Local Computer Use Agents
Release adds OTEL resource labels for slicing usage metrics, clearer agent task progress, hidden unused connectors, and resilient parallel tool calls that don’t fail as a batch.
v2.1.161 — What's changed OTEL_RESOURCE_ATTRIBUTES values are now included as labels on metric datapoints, so you can slice usage metrics by custom dimensions like team or repo claude agents rows now show done/total before the detail when work is fanned out; peek shows the longest-running item /mcp now collapses claude.ai connectors you've never signed in to behind a "Show unused connectors" row Parallel tool calls: a failed Bash command no longer cancels other calls in the same batch — each ...
The reactive Julia notebook reaches 1.0, signaling stability for reproducible, reactive notebooks and a maturing ecosystem around Julia’s interactive scientific computing workflows.
Pluto.jl 1.0 release – reactive notebook for Julia — Article URL: https://discourse.julialang.org/t/pluto-1-0-release/137296 Comments URL: https://news.ycombinator.com/item?id=48377496 Points: 19 # Comments: 0
New OSS interface for building and running coding agents ships with a polished UI and repo, aiming to standardize agent orchestration for developer workflows.
Show HN: Paseo – Beautiful open-source coding agent interface — Repo: https://github.com/getpaseo/paseo Homepage: https://paseo.sh/ Discord: https://discord.gg/jz8T2uahpH Comments URL: https://news.ycombinator.com/item?id=48377250 Points: 26 # Comments: 14
datasette-agent-micropython 0.1a0 debuts a WASM-backed MicroPython sandbox for safely generating and executing Python code; early tests show strong breakout resistance.
datasette-agent-micropython 0.1a0 — <p><strong>Release:</strong> <a href="https://github.com/datasette/datasette-agent-micropython/releases/tag/0.1a0">datasette-agent-micropython 0.1a0</a></p> <p>I want <a href="https://agent.datasette.io">Datasette Agent</a> to be able to generate and execute Python code safely. This alpha is looking promising so far. GPT-5.5 has so far failed to break out of the sandbox!</p> <p>Tags: <a href="https://simonwillison.net/tags/python">python</a>, <a href="https...
A Stanford Law blind evaluation finds leading AI systems outperform professors on key tasks, challenging assumptions about assessment, instruction, and tooling in legal education.
AI outperforms law professors in Stanford Law study — https://law.stanford.edu/wp-content/uploads/2026/06/salinas_... Comments URL: https://news.ycombinator.com/item?id=48377761 Points: 107 # Comments: 107
Kyle Daigle details GitHub’s approach to agentic development—platform constraints, roadmap, and coping with infra strain—as Copilot evolves from autocomplete to orchestrated agents.
GitHub's plan for Agents — Kyle Daigle, GitHub — GitHub pioneered the modern AI coding era with Copilot, and the resulting explosion in agentic coding has led to notable strains on the most popular developer platform in the world. Here's the plan.
