Golang/observability hacker changing the world @typesafeai. alum @docker @honeycombio @Bauplan_labs
San Francisco, CA
Joined October 2010
- Tweets19K
- Following3.5K
- Followers8.5K
- Likes37K
prompt engineering matters
Really excited to see @SrihariSriraman's chapter go live in "The Context Engineering Handbook." This is a killer systematic guide for creating and managing the core instructions for your agent or harness.
Nathan LeClaire retweeted
It has been interesting to see all the applications of @typesafeai's Jev at @sanity_io.
Like our new internal agent harness which allows anyone in the company to quickly spin up a bespoke agent. Jev is used for the router and has worked pretty well so far!
We (@rostimelk, @even, and I) wrote up an in-depth post about how it's built so you can do it too.
sanity.io/engineering/how-we…
Nathan LeClaire retweeted
1/ Agent-driven formal verification isn't a panacea. But it's damn close. And now we have the empirical data to back it up.
Using blinded agents, we compared formal verification against traditional code review. The results may shock you.
Still relevant prompting in 2026:
if a task is out of your expertise area, first go to ChatGPT and ask,
"Who are three real life people that would have good input on: <x>"
Then feed it into your agent "Do this as <person>"
They need something in latent space to latch onto.
paranoia is alive and well in the agent era. paranoia you'll get hacked, paranoia you didn't define the spec well enough, paranoia that your agent will leak your keys or delete prod...
alright that does it i'm building a subagent scheduler*
*markdown file on my computer i tell claude code to look in
alright guys you kinda played me encouraging me to be more subagentpilled
forked off like five and they all spun up jobs maxing out the cpus, causing a grinding twenty minute beachball fest
we need a local scheduler for agent workloads
typing "do it in a subagent", "do it in a subagent", etc. is getting really fucking annoying
should i tell my AGENTS.md to understand this to mean it?
go func() { make a worktree for that and do it }()
yes. i think this will be huge. jev is really good at telling if agent is progressing in the right direction. my early experiments with directing, guiding and correcting agents with it have been really promising.
Another insane way to use Jev:
Jev makes it dramatically cheaper to identify which corrections and trajectories from an agent run are reusable across agents.
And finally, someone open-sourced a self-improving memory layer that can put that signal to work across agent harnesses:
- Claude Code
- Codex
- Cursor
- OpenCode
- Hermes, and 20+ more
Beacon by @asymptotelabs continuously captures agent sessions across harnesses and stores their traces in a shared index.
I used it with Jev to build a live cross-agent memory workflow.
Beacon captures each run as it happens across all harnesses on my machine. Then Jev evaluates whether the trace contains a reusable engineering lesson, and the approved lesson becomes available to the next agent.
GitHub repo: github.com/Asymptote-Labs/ag…
(don't forget to star it ⭐)
Preserving an agent run and learning from it are two different problems.
Most traces contain file reads, failed commands, routine exploration, and implementation details that only apply to one task.
Passing the complete transcript to another agent adds context without identifying what matters.
In this workflow, one agent never receives another agent's raw transcript. It receives only the specific engineering lesson produced from the evaluated trace.
A correction made in one agent can then prevent the same mistake when another agent encounters a similar problem.
The recording shows the complete workflow I built.
If you want to dive deeper into Jev, I also wrote a hands-on guide to building this Jev-style decision path with open models, entirely locally.
Read it below.
Nathan LeClaire retweeted
1/ Our long national nightmare is over. The wall has been smashed. New community checkpoint for integer multiplication:
κ = 0.00268647 ≈ 2⁻⁸·⁵⁴⁰
T(n) = O(n(log n)⁰·⁹⁹⁷³¹³⁵³)
A 66% improvement since last announcement. A 2^173 fold improvement from the OAI result
*on a first date in san francisco*
so are you a queue-by-default or a steer-by-default in the agent kinda person?
Nathan LeClaire retweeted
Learn to build evals, folks.
Highly recommended reading on using decision models for evals.
From our internal evals, System One models like Jev are a great fit for LLM Judges.
If you want to go beyond Hamel's great write-up, I've also put together a few resources on applying decision models to LLM judges.
(bookmark it)
1) Jev-as-a-Judge for Agent Evaluations - academy.dair.ai/dashboard/re…
2) Jev-as-a-Judge: Evaluating Agents with a Decision Model - academy.dair.ai/labs/jev-as-…
Bonus paper I would recommend, with a few other ideas that combine System One and System Two models: Jev-as-a-Judge paper - arxiv.org/abs/2609.26550
