over the past month or so, i’ve struggled to find the time and headspace to write properly. i’ve been knee-deep in a new and exciting project, hopefully more on that soon. i’ve still been reading, but mostly things that didn’t feel like they belonged here. one exception was Shelby’s “The Sirens of Data Sales” , which deserves a thought of its own. her argument, as i understand it, is roughly this:…
give me an optimizable metric and i’ll move the world. benchmarking, i.e., the act of building and applying benchmarks, is a new(ish) form of data activation: a way of turning domain data into something models can be measured against, ranked by, and eventually trained on. i wrote about data activation a while back, and i want to come back to it, because i think benchmarking is one of the cleaner…
i am still pretty bad at collecting things i read. the attempts are worth it though, i think. mostly because the act of collecting pushes a little against the default “decay” (i.e., where links disappear into saved tabs, impressions flatten out, and then a few weeks later i remember only that something seemed “important”). this list is biased in the ordinary ways: some things stayed with me and i…
a nature medicine paper has been showing up in my feed recently: the one claiming general-purpose llms beat specialized clinical tools (openevidence, uptodate) on medical benchmarks. 1 i am having a hard time letting it go, so i want to write down why, as plainly as i can. the short version: they say they are measuring how good these systems are for medicine, and what they are actually measuring…
i read two agent papers a few weeks back, planning to write about both. by now one of them feels like a footnote while the other is the one i actually kept using. the one that survived is PEEK . PEEK maintains a small context map: a compact artifact that sits in the agent’s prompt and tells it how to orient inside a recurring external context. what is in this repo, corpus, or dataset? which…
my screen has looked the same for a few weeks now. on the left, codex (sometimes claude) in a main pane where i’m reading through a plan, asking why it picked this decomposition, sketching an html walkthrough of what the change would touch. on the right, a smaller pane handling things i don’t have to think about: a typecheck error in a file unrelated to the main work, a quick rename, a script to…
medmarks v1.0 dropped last week from the medARC group and got real attention, deservedly. it’s the largest fully open medical LLM eval suite to date: 30 benchmarks across a verifiable subset (Medmarks-V) and an open-ended LLM-as-a-judge subset (Medmarks-OE), 61 models on 71 configurations. eight of the benchmarks (collectively Medmarks-T) ship with train/test splits as verifiers environments,…
a quick one. small adjustments to how i work with agents, none of them deep on their own. but i can feel them adding up, the workflow shifting in a way that feels worth pausing on before any of it stops feeling new. that’s mostly why i’m writing this. the first change is that i’ve been trying to be more platform-agnostic. the reason is fairly mundane: i don’t want to feel locked in. i want to be…
a paper from Brodeur et al. (2026) has been making the rounds the past few days. 1 the framing it’s traveling with is “OpenAI’s o1 correctly diagnosed 78.3% of cases in NEJM clinicopathologic conferences, outperforming human physicians.” it didn’t click for me when i first saw it, and after sitting with it for a day i think i can articulate why, from the perspective of a (ex)physician and…
a week ago i wrote about a memory MCP i was using as a second brain . a few notes from week two. most of my actual work this week went into an evaluation framework i’m building, and the memory system was on for all of it (and for the rest of what was going on). three things i noticed. there’s a low-grade dread when i’m working: that no matter how diligent i am about notes, some lessons or threads…
i’ve been running a memory MCP as a “second brain” for the past week. a few notes, since the angle feels worth writing about even if i can’t tell yet whether it actually works. most of my work happens across a few different surfaces: Claude Code in the terminal, Claude Desktop on my mac, Claude mobile, a remote Claude Code session on an EC2 box, and now Codex for a second opinion (and i’d like to…
Borges has a very short story about an empire whose cartographers kept producing larger and larger maps, until they built one the size of the empire itself. 1 the following generations, less enchanted, saw it was useless and abandoned it to the weather. maps are useful precisely because they are reductions. when the pursuit of fidelity destroys the compression it destroys the point. for the…
so this post is comprised of a few thoughts that converged into a tool. first: building small things is the new procrastination. i acknowledge that fully. and yet i keep finding myself building small solutions for specific usecases, and i think it comes down to the fact that the famous xkcd is it worth the time has just shifted. it takes a lot less time and effort to build small tools you wish you…
For various reasons 1 , i found myself away from ‘work work’ for the past couple of weeks, and also mostly away from a laptop 2 , and because of that, first of all i feel rusty and out of touch with the research areas (and some aspects of agentic coding). we’ll see how to come out and fix that: i can double click on this feeling and say that a] it’s probably mostly ‘fake’ in the sense that stuff…
I needed a way to make diagrams for posts and papers. The requirements were simple: something I could mostly just ask Claude to create, something that looked decent for scientific work, and something I could edit afterward — not a PNG I’d have to regenerate from scratch every time I wanted to move a box. I looked at excalidraw and tldraw first, explored a few implementations, and ended up going…
A friend who drives a lot asked if there was an audio version of my posts. There wasn’t. So I looked into it — and it turned out to be absurdly easy. The whole thing took an evening. That’s the part worth writing about. Open-source text-to-speech crossed a quality threshold sometime in 2025, and I completely missed it. Kokoro-82M 1 is 82 million parameters — tiny — Apache 2.0 licensed, and it…
I’ve been noticing something about how I work with Claude. It’s not dramatic — no moment where the model led me off a cliff and I realized too late. It’s more of a drift. The models have gotten good enough that my default reaction to most outputs is “yeah, that’s probably right.” And most of the time it is. But “most of the time” is doing a lot of work in that sentence. The issue isn’t that I’m…
Opus 4.6 came out ~20 hours ago and I wanted to get a feel for it. It also seemed like a good chance to follow up on some of the ideas from my previous post — specifically, building small tools that help me stay more engaged with my own work rather than just producing output. 1 Opus 4.6 The experience isn’t wildly different from Opus 4.5, but the edges are smoother in ways that matter. It picks up…
How can someone who enjoys thinking — enjoys the cognitive load — use coding agents and LLMs to foster continued learning, and not skill degradation? And what are some useful mental frameworks to have in mind? I think this is really about the reality we all live: the reality of coding agents and chatting with LLMs and how it affects us building (or eroding?) our skills. I always considered myself…
In the last couple of months we’ve been working on a health agent. It was my role specifically to deal with the messy answer to the tough question: “is it good? worthwhile? valuable?” I’ll try to describe here our attempt to answer this question. This follows Anthropic’s Demystifying evals for AI agents , which I recommend reading first if you haven’t. 1 What I’m adding here is the specifics of…
“The Entertainment is real, and it’s called Instagram Reels.” — Will Gottsegen, The Atlantic Not a unique observation (the connection is clear). But there’s something I like about people still arriving at Infinite Jest and feeling compelled to say it out loud.
A foundation model I’ve been working on recently got published in Nature. 1 For a while I’ve wanted to write this. Now the paper is finally out so I have to do it in a timely manner, and I also have to start investing more thought in the upcoming projects (some are similar). So what is this? I think the honest answer is something between a post-mortem of a successful project and some exploration…
The landscape is shifting in recent years — it’s a cliche to start texts like this these days, but the fact that it’s a cliche doesn’t make it any less true. 1 In 2019, the folks at Andreessen Horowitz wrote this about data (in a piece titled The Empty Promise of Data Moats ): “Instead of getting stronger, the defensible moat erodes as the data corpus grows and the competition races to catch up.”…
For a while now I’ve been collecting too many things to read and think about, mostly in twitter ‘saved links’. This is a place for me to silently collect together things I enjoyed reading / other content I’ve enjoyed and sometimes jot some thoughts about it. It’s built in a way that won’t make me feel too committed, should be kind of under the radar, low friction, minimal effort.