iiinova — where your digital self keeps evolving

Arthur Pastel

@apas.tel

Co-founder @codspeed.io • Building trustable and consistent performance tooling • Python, Rust, TypeScript • Creator of ODMantic • apas.tel

Posts Replies Media

Arthur Pastel @apas.tel · 03/08/2026

CodSpeed v5 is out, with two big changes to how we measure your code. First, cycle estimation: every instruction is now weighted by its measured cost on real CPUs, cache misses included. A `div` no longer counts the same as a `mov`.

CodSpeed v5: Only your code, at its real cost

100

Arthur Pastel @apas.tel · 29/05/2026

"I use it every day and honestly can't live without it. Saves me tons of time." hardfist, lead of @rspack.dev at ByteDance, on the CodSpeed Wizard. So excited to build the performance agent for the fastest projects on the planet!

010

Arthur Pastel @apas.tel · 28/05/2026

"If you're gonna ship slop, at least make it fast." Our top sticker at PyCon and there's a reason: code has never been cheaper, reviewers are under pressure and regressions are piling up. Stop thinking about performance, we'll make your thing faster.

001

Arthur Pastel @apas.tel · 26/05/2026

Claude Code can now improve performance and fix regressions. The CodSpeed plugin just landed on the official marketplace. Try it out: /𝚙𝚕𝚞𝚐𝚒𝚗 𝚒𝚗𝚜𝚝𝚊𝚕𝚕 𝚌𝚘𝚍𝚜𝚙𝚎𝚎𝚍@𝚌𝚕𝚊𝚞𝚍𝚎-𝚙𝚕𝚞𝚐𝚒𝚗𝚜-𝚘𝚏𝚏𝚒𝚌𝚒𝚊𝚕

101

Arthur Pastel @apas.tel · 05/06/2025

Announcing p99.chat, the assistant for software performance optimization 🚀 Describe your performance challenge, share your code, and p99 will analyze, instrument, and visualize the results, cutting the traditional setup overhead down to seconds!

141

Arthur Pastel @apas.tel · 30/05/2025

We’re at @pyconit.bsky.social with @adriencaccia.bsky.social! We’ll quizz you on Python performance again tomorrow! See you in room lasagna at 11:45!

010

Arthur Pastel @apas.tel · 04/03/2025

Super excited about our latest changes to make performance testing more accessible! Two significant changes: ✅ The Free plan now includes private repositories (up to 5 users) ✅ Macro Runners are now globally available (bare-metal ARM64, 16 cores, 32GB RAM) with 120 min/month included for free

100

Arthur Pastel @apas.tel · 28/02/2025

We now support sharded benchmark execution to provide even faster performance feedback!

121

Arthur Pastel @apas.tel · 19/02/2025

But physics is catching up. As transistors shrink, heat and power constraints limit clock speed gains. But to keep Moore’s Law going, we stack more cores. GPUs push this further with thousands of parallel cores, mentioning exponential performance increase but omitting the implementation complexity.

100

Arthur Pastel @apas.tel · 19/02/2025

Single-core performance gains have largely plateaued. That means no more relying on “free” hardware speed boosts. CPUs haven't stopped improving, but we must actively tap into new architectures. Concurrency, SIMD, and thoughtful optimizations can still unlock impressive speedups.

100

Arthur Pastel @apas.tel · 17/02/2025

`x * x` generates exactly the same assembly as `x.pow(2)` in Rust and this goes even beyond powers of 2, thanks to LLVM loop unrolling and exponentiation by squaring.

110

Arthur Pastel @apas.tel · 12/02/2025

Started hacking on a small TUI for managing EC2 instances and SSH connections. Can’t wait to ditch AWS Console's clunkiness.

100

Arthur Pastel @apas.tel · 10/02/2025

Lets go! I’ll be in Pittsburgh again in may for PyCon US 2025! Super excited to try a new interactive talk format this year!

000