Max Ghenis
banner
maxghenis.bsky.social
Max Ghenis
@maxghenis.bsky.social
600 followers 41 following 82 posts
Co-founder and CEO of PolicyEngine
Posts Media Videos Starter Packs
Heading to LA this week for Benefit Navigator's event.

Their platform helps families and case managers navigate public benefits and tax credits - powered by the @us.policyengine.org API for tax and benefit calculations. Excited to see tech and policy come together to expand access.
Tomorrow evening I'll be celebrating the 13th anniversary of the @opengovhub.bsky.social, @us.policyengine.org's new home in DC. I'll also give a short talk on our work to promote open governance and transparency, the mission of the space and our neighbor organizations. Tickets still available!
13th Anniversary Celebration of the Open Gov Hub
Celebrate 13 years of Open Gov Hub!
www.eventbrite.com
ChatGPT estimates the anti-Israel boycott against Shouk increased their risk of closure by 40% this year.

Links:
instagram.com/p/DHWA8BWJHl6
washingtonsocialist.mdcdsa.org/ws-articles/...
instagram.com/p/DPUllyAjkxb
chatgpt.com/share/68e223...
Disappointed that the plant-based Shouk, my favorite fast-casual place in DC, has closed. Anti-Israel groups targeted it because an Israeli immigrant owned it, and are now cheering their victory.
Agentic AI coding tools continue to inspire us to turn everything into software.

Join us at our new home in downtown DC, the @opengovhub.bsky.social, for a celebratory happy hour on Oct 22: luma.com/lnrzph2h
Alaska's Fairbanks region has the highest second-lowest-cost silver plan (a key PTC ingredient) for older residents.

As a result, a married couple, each 64 years old earning $83k, would gain $57,292 in 2026 if the enhanced subsidies are extended. They'd get some PTC with up to $756k income.
Our new ACA calculator shows how extending the enhanced premium tax credits would affect households.

An extreme case: A 64-year-old in Fairbanks, AK would gain $655 in 2026 if they earn $61k, but gain $26,861 if they earn $62k.

Try it here: policyengine.org/us/aca-calc
Come by tonight if you're in London!
Economic policy drinks tomorrow in London 🍺

Thursday 6-8pm at The Sherlock Holmes (Northumberland St, near Embankment)

Light bites + first round on PolicyEngine. Come meet economists & policy analysts working on UK tax/benefits/public finance.
luma.com/oz1qt2ec
Economic policy happy hour · Luma
Drinks and discussion for the UK economic policy community. PolicyEngine is hosting an informal happy hour for economists, researchers, and policy analysts…
luma.com
Can't wait for our first live event in London! If you're in the area and interested in policy simulation, I think you'll discover major new possibilities our new app and research will offer.

See you there: eventbrite.co.uk/e/policyengi...
We constructed fully synthetic firm microdata to conduct this analysis. I was struck how calibrating to a large range of coarse aggregates could result in microdata that reveals sharp responses to cliff effects.

You could run a bunching study on this synthetic data! (And we will.)
This partnership exemplifies why we built PolicyEngine: policy analysis tools should be open, accessible, and sustainable.

Honored that NBER trusts us to ensure TAXSIM—the foundation of tax research for decades—continues serving researchers while evolving for modern needs.
🤝 Big news: PolicyEngine and @nber.org
have signed an MOU to ensure researchers worldwide continue accessing critical tax microsimulation tools.

We're building an open-source TAXSIM emulator, combining NBER's 50+ years of tax expertise with modern open-source infrastructure.
Here I also used LLMs to check my intuition on how readers might interpret articles (all four I tested interpreted the Telegraph's statement the way I thought the public might; that is, incorrectly).
HMRC projected last year that raising the VAT threshold to £90k would increase revenues in 2028-29. Looks weird until you realize that would represent a tax hike by then, since it'd otherwise be £92k!
Today, @telegraphnews.bsky.social suggested that raising the UK VAT registration threshold would increase revenues in the long run. This misread (poorly documented) government projections, which reflected inflation adjustments in the baseline. My latest:
substack.com/home/post/p-...
VAT thresholds, revenues, and the role of counterfactuals
No, the OBR has not projected that raising the VAT threshold would increase tax revenues.
substack.com
We have a suite of subsequent checks (separate agents) after test-creator and rules-engineer complete. If they misalign that gives more to interrogate.

We'd like to test our hypothesis here with a controlled experiment, but my guess is that our approach reduces cascading errors.
Tests can be wrong! This way we get two interpretations of the specifications (statutes, regs, manuals), coming from different perspectives. If the agent misinterprets the spec in constructing the test, it's more likely to make the same error at the encoding stage--less so if they're starting fresh.
I see this as the next iteration of test-driven development, free from the biases you can have by writing the code after you've already conceptualized its logic by writing tests. We expect separating the work to improve quality.

Full video here (I started at 4:46): youtu.be/vss_slgfzIg?...
Modeling Texas LIHEAP with AI Sub-Agents | PolicyEngine Demo
YouTube video by PolicyEngine
youtu.be
I've been using Claude Code a LOT lately ($5,800+ in tokens last month on my $200 Claude Max plan), and now experimenting with custom slash commands and subagents. Here's an example: using parallel subagents to independently write integration tests and encode policies for @us.policyengine.org.
Reposted by Max Ghenis
Bumping the cap to $62k/124k would benefit 22% of US residents in the top income decile and 1.2% of those in the bottom 90%.
The SALT cap would raise taxes on 22% of Americans in the top income decile, and 3% of Americans in the bottom nine deciles.