faraz

@LucasPCaccia 1000% agreed

@farazdotai I don't think this is false, but I do believe that you accumulate a debt by not having a deep understanding of the core components of your code. You save time today but you pay it tomorrow

Witnessed two interns go through a series of graphs, analyze the ops and shapes of tensors like by line Admirable they take the time to learn, but claude could have made a report and explain in 0.001 of the time

Man the early days of ChatGPT were so special

@SinaHartung founders are wealthy not rich (on paper)

startup founders are just poorer forward deployed engineers

@subminima Nothing wrong w it, just had not witnessed it in a while

@farazdotai outsourcing understanding?

@kennykgguo @evanliin we spent 1 hour analyzing it together afterwards

@farazdotai lol this was @evanliin and I

June ICU 👀

May was tough on the budget for sure

The best harness is no harness

Jk Codex is the real one writing, I just read markdown

Crazy how 98% of the kernels I write are in pythonic declarative language level. Perf tradeoffs exist, but Python+JIT is goated.

@alish2001_ Bro gets premium cuts

@farazdotai +60$

May expenses living in Toronto - Rent $2550 (parking incl) - Utils & Internet $140 - Phone+Plan $106 - Insurance $280 (car+condo) - Groceries $516 - Subscriptions $127 (ai/yt/spotify, x) - Gym $77 - Eating out: $320 - Gas $115 - Shopping $400 $4631/month +$5355 furniture

new grads often ask me what they should be doing so they don't fall behind in the ai space. there's a lot, but its honestly super manageable. become intimate with model internals. proof based linear algebra. non-convex optimization. this is stuff you could've done in undergrad. it definitely takes some time and work, but its doable. have taste, have opinions. train a small model, then train a big one. vLLM internals, tensor parallelism. hand roll kernels. cluster orchestration. do you have opinions on synthetic data? why don't you? SFT, PPO, you should know this. learn Triton. everyone is reproducing papers now so you need to be doing more. do you know the semi supply chain? where are the bottlenecks? hardware, man, hardware. your little gpu rig erector set in your basement isnt gonna cut it. build a cluster, a big one. pretrain a 800B model. now postrain it. serve it to millions of people. you should be able to beat deepseek on some benchmarks now. its a lot to take in but it all snowballs. this what job security looks like from now on. do you want to work in tech or not

@peteoxenham Claude find the bottleneck in my training split

strava mcp launched today lfg

Used to wake up to check the market/socials, now I check my agents' aggregate reports. Andrew Huberman, you did not see this.

@serenaa_ge This is sick, you go Serena

Today we’re releasing DeepSWE, a new standard for agentic coding benchmarks. On public leaderboards, top models often look relatively close in capability. DeepSWE shows where they actually diverge, reflecting the realistic experience of developers in their day-to-day work.

@Yuchenj_UW Couple background agents running with auto prompting for follow ups can easily reach 300M tokens a day

An OpenAI friend told me he burns 300M GPT-5.5 tokens/day. The top one in his team burns billions of tokens/day. Codex coding for them every night. Databricks also gives engineers unlimited tokens. We're looking for cracked inference engineers to join us at Databricks AI to produce trillions of tokens, insanely fast. DM me if you have: - Contributed to open-source ML systems like SGLang/vLLM/PyTorch - Experience serving LLMs at large scale Databricks AI runs like a startup. Lots of exciting things to build!

@pxue Waterloo new grad from last year here, I don’t think salary alone is the right measure at such early stage. At Waterloo, we value growth (including compensation growth) more than temporary economics.

this kind of comp in Toronto gets me excited. $130k salary you can rent a $2.5k/m unit downtown for about 25% of your gross salary, so no commute time. comfortably saves if you want to, and get meaningful equity ~10k options vest over 4 years. i'd like to see new grad out of Waterloo get this role role.