Samuel Colvin

@GergelyOrosz LG UltraFine 40U990A Changed my life (in a small but very satisfying way).

My current work monitor has issues, so looking for a new one. I could get a replacement Dell 27-inch one for ~$600, or an Apple Studio XDR 27-inch for $3,000. I was thinking: would this 5x as expensive monitor help me do better work *at all*? I see parallels with AI spend...

A Pydantic AI agent on your laptop is easy to reason about. Production is less forgiving: a pod dies, a tool times out, a human's offline for approval. New guest post from @htahir1 on Kitaru, a new runtime layer underneath Pydantic AI 👇 pydantic.dev/articles/runti…

@mitsuhiko okay, that's enough for me. I'll stay away.

@samuelcolvin Skits about human <-> machine relationships and machines being equated to slaves. The narrative that unfolds then is a lot of and blacks and apartheid tropes.

After reading some comments on Hackernews I decided to download TikTok and search for clanker. There is definitely a certain … vibe to the type of content there. Fascinating.

@devagrawal09 @antisymmetrics @AAAzzam @oriSomething @pydantic Are you sure that's true when LLMs are writing the code?

@antisymmetrics @AAAzzam @oriSomething @pydantic @samuelcolvin the best type systems include an effect system

typescript is the wrong language for code mode agents

@devagrawal09 @AAAzzam @oriSomething @pydantic Models are trained on hundreds of millions of lines of python. Good luck ever promoting them to do as well with a new language.

@devagrawal09 @AAAzzam @oriSomething @pydantic Try it. You'll need thousands of tokens to explain how to write the language. With Monty you say "write python with some limitations" a few hundred tokens, hopefully one day, 2 tokens.

Loving github today.

This is peak 2026 - 180 retweets, 1.3k likes, 1.2 bookmarks, not a single issue on the repo. So I looked at what's actually in here. I was not impressed. Since none of you will go to github anymore, here's a screenshot of my questions. github.com/workos/auth.md…

@charles_irl Don't agree I'm afraid. The slides CAN look great. And you CAN collaborate with the AI to write and learn good content.

wild to me that people vibe-generate slides for conference talks they are ugly (for now). they are low info densiry (thanks rlhf) but worse, they don't represent your thoughts, so your presentation of them will be terrible, unless you put in a ton of work (so just write them!)

If you showed them this tweet, they would nod. But when many of them don't know the difference between MCP and an agent framework, it's hard for them to identify what solution are actually valuable, so they go back to finding ways to get anthropic and openai shares, since that seems like it's not a bad idea.

@samuelcolvin I agree with this assessment. It is my experience with Elastic customers too. I just didn't know VCs would ignore this.

ARR is cool because it's the last resort of the blind. Speaking to VCs at the moment, and the profound ignorance of many of them about how AI is actually being used is striking. The assumption is they're "all in", they're not - most VCs are doing the consensus thing because they know they don't know better.

Not sure I have a compelling one-tweet answer to that question, but here goes: AI is incredibly powerful, but a massive engineering headache. So there are three stages: 1. realise it's incredibly powerful 2. realise it's a massive engineering headache 3. figure out the engineering headache to make it useful Most companies are someone in step 1 or 2. Very very few have completed 3 (except for "use chatgpt/claude as much better google search"). There's enormous TAM in helping with step 2, but we're early.

@samuelcolvin Out of curiosity, what are they missing about how AI is actually being used?

@JordiMonPMM I think it's a fine financial metric. But if VC was only about measuring financial metrics, any fool could do it. It's surely hard to defend 2%+20% fees if your job could be done by anyone charming with a spreadsheet.

@samuelcolvin What other financial metric do you think is more relevant, rather?

@nontrivialzeros Great news!

@samuelcolvin We’re good :) Nice to meet you at PyCon!

Really hoping the pyrefly team weren't affected, anyone know if they were? 🤞

🤦 dyslexia! * bored

I got board of opening spreadsheets with google sheets. So I've built `spread` a native spreadsheet viewer. Built with @zeddotdev's GPUI. Can open xlsx, csv and parquet. Will open a 30M row parquet file in <100ms. Renders most xlsx formatting. github.com/samuelcolvin/s…

@charliermarsh To think we had to work so hard to avoid a ~5% perf regression, when all these fruits were hanging below us.

/goal for finding silly one-line optimizations that speed up your parser by 20-30%

What's the definition of performant software? Answer: @codspeedhq

@dctanner @zeddotdev I totally agree. Please don't nerd-snipe me into building that too. I have a job!

@samuelcolvin @zeddotdev The other thing I would love (but probably is best in Zed itself) is a rendered markdown editor, aka Notion but for files on my computer. The future of docs and sheets is .md and .csv