Toronto, Canada
Joined November 2008
- Tweets12.1K
- Following1.5K
- Followers4.8K
- Likes4.2K
Told a civil engineer that AI writes all my code now. His jaw dropped. I wonder if he saw his own job drastically changing in the same way as mine has or did he realize programmers might someday be obsolete?
Pretty happy to see this. Lots of good memories in this building
obj.ca/shopify-signs-new-lea…
Love the momentum around System One models
a brand new model paradigm! introducing 𝚌𝚕𝚊𝚒𝚛𝚒𝚌: a system one model for extracting, computing, estimating, and predicting values from unstructured inputs.
blog.cloudflare.com/clairic-…
Starting to wonder if “pair programming” with AI could ease the PR review bottleneck
Gathering context makes people say, “I’ll review this tomorrow.”
🤖 Made with AI
I think a year from now it’ll feel quaint that we were editing SOUL.md just to coax a bit of dry humour out of our coding agents
I feel like we need to vibe-code one of these
🤖 Made with AI
I’ve been working on a bunch of AI initiatives at @Opendoor. Hard problems. Lots to build
If you’re in Toronto and curious, come chat with me, @MargaretSyms, and @EyalKlein1 and others on October 7
Limited spots: docs.google.com/forms/d/e/1F…
Jev can turn past LLM responses into a semantic cache: when a new request has the same intent, reuse an answer instead of paying to generate it again. That feels profound. It isn't just faster and cheaper. It could help less-funded companies build AI agents
I feel like we need to start building some tradeoff escape hatches so we can move faster. This technology was created to help us move fast, but I think we’re trending in the wrong direction. It’s the right direction for normies, but the wrong one for people who know what they’re doing
Do you want to produce slop quickly or slowly?
Quick slop: Let an LLM build the entire codebase
Slow slop: Write small, tidy PRs, then watch AI agents gradually turn the codebase into slop through their reviews
You have only one choice. Which do you choose?
This discussion was refreshing
Jev and the System One Model: RLCD, intelligence/$, reliable AI, & the end of chat-first AI latent.space/p/jev
@typesafeai CEO @CompleteSkeptic explains why AI can solve extraordinarily hard problems yet still fail to automate basic work, why Jev is built for reliable decisions inside software instead of chat, why TypeSafe rejects public benchmarks and refusals at the API layer, why data and the right task matter more than brute-force compute, how System One Models could reshape coding agents and software, and why even with $1 billion he wouldn’t pre-train a model from scratch.
The most exhausting AI loop: watching two agents fight over how to clean up a PR while you bite your nails hoping they don’t expand the scope or make the code worse
HA. Tried calculating lat + long with Jev. Ten criteria isn’t enough for regression, but it could make an interesting eval for whether Jev understands locations
