MacKenzie Scott has given away $26 billion. What did it buy in health?
An interactive model from her own gift database: ~70,000 QALYs weighting each study by causal credibility; ~200,000 at face value. Drag the assumptions yourself: maxghenis.com/mackenzie-sc...
Social feed
All my posts from across the web, aggregated in one open source location. Bluesky posts update at build time. X and LinkedIn posts are imported from exports.
straude.com/leaderboard?...
I just became the first Straude user to report 1 billion output tokens.
(The dollar column is API-list-equivalent — I'm on subscriptions.)
Most of them went into things we start shipping this week.
For a whole year! 1 beer a year
To be fair I probably do a few hundred Claude Code requests a day but I also don't drink
Happy July 4! In case you were wondering about the water footprint of that beer you're holding, I did the math: 53,300 ChatGPT queries.
Running 10 Claude Code tasks a day for a full year? Also one beer.
Drink responsibly. Prompt freely.
maxghenis.com/drinking-ai
This appears to be the first official U.S. federal social-science statistical release to rely on an external open-source model.
Last night at Civic Builders DC. Tax and benefit rules are public — computing them isn't. @policyengine.org makes that computation free and open source: change any reform, see who wins and loses, down to your own household.
Slides: policyengine.org/slides/civic-builders-2026
The US government singled out one AI lab — by administrative letter, under no published standard — and cut roughly a third of its own researchers off from the frontier models they build. The same researchers could use internal frontier models at any competitor tomorrow. substack.com/home/post/p-...
We built PolicyBench to test whether frontier models can calculate a household's taxes and benefits without tools.
The best, GPT-5.5, gets 80.3% to the dollar.
To get to 100%, agents need to hand this to a deterministic engine. Just too hard to map out the statutes, regs, and forms in your head.
Spent last week at @rockefellerfdn.bsky.social Bellagio with leaders from government, technology, and philanthropy, on human-centered government in the age of AI. The line that survived every argument: the system, not the person, should carry the burden of complexity.
My slides: policyengine.org/slides/pwbm-2026
Presented @us.policyengine.org at @budgetmodel.bsky.social — rules, data, behavioral assumptions, and where AI is taking each. Thanks to Kent Smetters and the PWBM team for hosting.
This raises a question worth taking seriously: what if every bill could have a score, on every indicator?
My slides: policyengine.org/slides/cbo-2026
Gave a talk at the Congressional Budget Office this week on AI and policy modeling.
A policy projection rests on four pieces: encoded rules, calibrated data, dynamic responses, and the analysis that turns numbers into counsel. AI is now reshaping each one.
I built OpenMessage to give Claude Code / Codex access to my Google Messages texts. Recently added WhatsApp and Signal support, and I now use it daily instead of those three native apps. Give it a whirl and let me know what you think.
maxghenis.com/blog/talkie-...
I tested Talkie-1930's math abilities.
On rote arithmetic, it matches or exceeds GPT-3 175B (13× larger).
On 5-shot GSM8K word problems, it scores 4.9% — same-size LLaMA 13B scores 17.8%.
Also built a tool to explore its responses.
Looking forward to this, especially:
- Panel with three orgs, three products, same open rules engine
- @daphnehansell.bsky.social's Coverage Compass talk
- @arielmai.bsky.social's Rules as Code lunch
If you're in Chicago let's connect!
My slides: www.policyengine.org/slides/eagxd...
AI will change society more quickly than ever before, and policymakers will have to respond with unprecedented urgency. We need to equip them with the most predictive combination of economic models, forecasting, and AI to align outcomes to our shared values through policy.
This is an example X post. Export your X archive and run the import script to populate this file with your real posts.
This is an example LinkedIn post. Export your LinkedIn data and run the import script to populate this file with your real posts.
How to update
This feed combines posts from multiple platforms. Here's how to keep it updated:
- Bluesky: Updates automatically at build time via public API
- X/Twitter: Request your data archive from X settings,
then run
npm run import-xto convert it - LinkedIn: Export your data from LinkedIn settings,
then run
npm run import-linkedinto convert it