CodSpeed

New flamegraph feature just dropped!

A benchmark can move without a single line of your code changing. A compiler upgrade. A new system library. A different CPU config. Until now, telling that apart from a real regression meant digging through CI logs. CodSpeed now diffs the environment between any two runs: toolchain versions, linked libraries, CPU flags. Identical values collapse, so you only see what actually changed. The same diff ships in the MCP server's compare_runs report, so your AI assistant factors it in when investigating a regression. On by default. Nothing to configure. Learn more: https://lnkd.in/eGqn2s9t

OpenAI, Cloudflare, Vercel...  all running on CodSpeed. And they're now hiring their founding Growth 🔥 When I met Arthur Pastel and Adrien Cacciaguerra in SF, it took about ten minutes to see they were onto something. CodSpeed catches performance regressions before they ship. Developers discover it through performance reports on open-source repos, adopt it on their own projects, and it spreads from there. Hundreds of open-source projects and the enterprise teams behind them were running it before anyone ever reached out. This is not a quota-carrying AE role, but a real GTM engineer who owns pipeline generation end-to-end. Turn usage signals into qualified pipeline, find the developer who already loves the product and arm them to sell it internally. If you've done Growth or GTM engineering at a developer-led product (Vercel, Sentry, PostHog, and the like) and want to build the motion from scratch this time, this is one of the best seats I've seen this year. SF or US preferred. JD in comments.

If you can't measure performance, you can't optimize it. Not as a human, and definitely not as an agent. Last week at Neon Noir., I showed a fix where my intuition was wrong, the room's was wrong, and LLMs were wrong. Three extra instructions, zero cost. That's the whole problem with letting a model "optimize" your code. It guesses with confidence and ships regressions. An optimization agent is only as good as the measurement underneath it. That's what we're building at CodSpeed. Thanks Max Corbani for organizing this great event, to Tom Houlé (Zed Industries), Steeve Morin (ZML) and Ruben Fiszel (Windmill) for the talks and to everyone there for the great conversations!

Code. Infer. Commit. The lineup is set. → Tom Houlé (Zed Industries), building the code editor for the agentic era. → Steeve Morin (ZML), shipping high perf inference on any hardware. → Ruben Fiszel (Windmill), scaling orchestration without breaking it. → Arthur Pastel (CodSpeed), catching perf regressions before they ship. 250+ builders already RSVP'd. Only a handful of seats left. June 18th. Neon Noir. Open source, beers & pizza. RSVP in the comments.

CodSpeed a l'air super intéressant ! Je l'ai découvert en enregistrant un podcast pour #IFTTD et c'est maintenant dans ma liste des outils à tester ;)