←
mufeez668 posts
mufeez avatar

mufeez

@moofeez

having fun exploring ML infra @linear, fmr @meta. @uwaterloo SE alum

◷ Joined December 2015▧ 86 Photos and videos
mufeez avatar
mufeez@moofeez·Apr 28

I post-trained Qwen3-Coder to fix bugs using an actual debugger. The result: Solve rate: 70% → 89% Median turns to fix: 46 → 19 (-59%) Instead of just reading code or print-debugging, it: - reasons from execution - inspects live variables and call stacks - sets breakpoints, steps, and evaluates expressions

0:38
mufeez avatar
mufeez@moofeez·Nov 9

do people actually publish their env files like this...?

Post attachment
mufeez avatar
mufeez@moofeez·Sep 30

Someone should study the correlation between “Labs” in the name and dollars raised

mufeez avatar
mufeez@moofeez·Sep 27

It's been 3 weeks since Muse launched but I'm still using Instinct, am I missing out?

Tom Moor avatar
⟳ mufeez retweeted
Tom Moor@tommoor·Sep 21

It passes the test! Send this post to your agent and watch it work wonders on your CI times – also read and learn ;)

Linear@linearSep 21

Agents were shipping code faster than our CI pipeline could keep up. Our team optimized our pipeline, leading to a roughly 15% faster test suite and 50% reduction in runner-time spent per test, all while our test suite grew 4x in size. @moofeez explains how: linear.app/now/ci-bottleneck…

mufeez avatar
mufeez@moofeez·Sep 21

Our test suite is growing 9x faster than a year ago, now that agents write most of it. If we'd left Cl alone, a PR would wait ~11 minutes today. We got it to 5, on 3.7x the tests. Give the blog post to your agents for free gains!

Post attachment
Linear@linearSep 21

Agents were shipping code faster than our CI pipeline could keep up. Our team optimized our pipeline, leading to a roughly 15% faster test suite and 50% reduction in runner-time spent per test, all while our test suite grew 4x in size. @moofeez explains how: linear.app/now/ci-bottleneck…

Tom Moor avatar
⟳ mufeez retweeted
Tom Moor@tommoor·Sep 18

The test of a great technical blog post at this point is: can I point an agent at it and have the learnings translated across to my codebase?

mufeez avatar
mufeez@moofeez·Sep 17

really not liking the trend of code review becoming agent ping pong…

mufeez avatar
mufeez@moofeez·Aug 27

btw make friends with your AWS/GCP rep now if you want compute next year

mufeez avatar
mufeez@moofeez·Aug 27

imo Instinct worth at least 5B after they implement this

mufeez avatar
mufeez@moofeez·Aug 27

how long until my Instinct can talk to your Instinct directly

mufeez avatar
mufeez@moofeez·Aug 26

Recently crossed 3 years at Linear Wild how far a cracked team + beloved product takes you 🫡 (we're hiring!)

Linear@linearAug 26

Today we're announcing our second employee tender offer. We passed $100m ARR earlier this year and now have more than 40,000 paying customers. The tender lets our teammates participate in that success at a $2.5B valuation: linear.app/now/sharing-growt…

mufeez avatar
mufeez@moofeez·Jul 27

Kimi K3’s RL setup is surprisingly “boring” synchronous RL, specialist policies, MOPD maybe a great pretrain is all you need

Releasing the model weights and technical report of Kimi K3. Kimi K3 is our most capable model: a 2.8T MoE model with native visual understanding and a 1M-token context window. New model architecture: 2.5x the intelligence per unit of compute, not just more params. Alongside Kimi K3, we're opening up more of the stack behind it — high-performance attention kernels, MoE communication library, and infrastructure for running agent environments at scale. Model weights: huggingface.co/moonshotai/Ki… Tech report: github.com/MoonshotAI/Kimi-K… Tech blog: kimi.com/blog/kimi-k3

Post attachment
mufeez avatar
mufeez@moofeez·Jul 24

ending the week with a ~36% drop in median CI time 🫡

Post attachment