X

@poteto

Volume does matter: article arguing that, with agents, PR count and volume become a meaningful signal of output; responds to a question on why anyone ships hundreds of PRs a day. post (https://x.com/poteto/status/2108290818746528017) · 2,843❤ · 156 RT

Be more ambitious: short post urging engineers to be far more ambitious. post (https://x.com/poteto/status/2108299315764777171) · 7,107❤ · 650 RT

Grok Bot on Omarchy: asks Omarchy users and plugin builders for feedback and feature requests on Grok Bot integrations. post (https://x.com/poteto/status/2108261735199371451) · 1,327❤ · 43 RT

@cursor_ai

/visualize follow-ups: you can ask a follow-up question in the same chat and /visualize answers with a new chart. post (https://x.com/cursor_ai/status/2108289748628566496) · 974❤ · 64 RT

@hwchase17

Evals and Goodhart's law: eval-driven development works for narrowly scoped tasks but not for more autonomous agents; asks how to hill-climb those. post (https://x.com/hwchase17/status/2108474622966444102) · 41❤ · 4 RT

Trajectory labeling: labeling is several questions, not one pass/fail; Jev as a judge in LangSmith evals gives difficulty and correctness separate typed answers per trace. post (https://x.com/hwchase17/status/2108453917277254058) · 37❤ · 3 RT

Decision models in a harness: shares a First Pass interview with LangChain's Sam Crowder on where decision models fit and how to use Jev with LangChain. post (https://x.com/hwchase17/status/2108272060141301908) · 27❤ · 8 RT

@simonw

Microsoft sandbox library: new open-source cross-platform (Windows, macOS, Linux) agent sandboxing library built on processcontainer, bubblewrap, and seatbelt; also notes Microsoft Quicksand (QEMU in Python) and AWS Strands Box. post (https://x.com/simonw/status/2108216753604248000) · 900❤ · 78 RT

"Super Intelligence" naming: criticizes the use of "Super Intelligence" as a label, citing a DOJ headline about AI-assisted streaming fraud. post (https://x.com/simonw/status/2108346075874734127) · 1,152❤ · 54 RT

@swyx

Hiring niche: quotes a post on the hard-to-hire AI media/technical operator role, stating comp is $5-50M at frontier agent labs. post (https://x.com/swyx/status/2108379841859072028) · 601❤ · 9 RT

@rauchg

Discovering talent: says spotting greatness in people early is one of the most enjoyable parts of work. post (https://x.com/rauchg/status/2108399406009798764) · 817❤ · 29 RT

@HamelHusain

Same model as judge: usually fine to use one model for task and evaluation, as long as you test judge agreement with human labels on a held-out set. post (https://x.com/HamelHusain/status/2108210834988032433) · 22❤ · 5 RT

Silent

@martinfowler, @karpathy: no original posts in window

Gaps

@xAI: API not authorized to read this account's posts (skipped)

GitHub trending

boykopovar/AnyPS5 (https://github.com/boykopovar/AnyPS5): tool for automatic porting of PS5 executables to Linux and Windows (C++, +4,669★ today)
cathrynlavery/diagram-design (https://github.com/cathrynlavery/diagram-design): editorial diagram design for coding agents, 42 diagram types as self-contained HTML + SVG (HTML, +1,160★ today)
morluto/rea (https://github.com/morluto/rea): reverse engineer app behavior down to native binaries with agents (TypeScript, +7,738★ today)
mattpocock/skills (https://github.com/mattpocock/skills): skills for real engineers from his .agents directory (Shell, +1,774★ today)
thedotmack/claude-mem (https://github.com/thedotmack/claude-mem): persistent context across agent sessions via captured, compressed, re-injected memory (TypeScript, +670★ today)