Daniel Milo

@mycelias On the flip side, governments may use AI themselves to write longer and more complex regulations/processes. Still unclear where the equilibrium will be.

it seems like legal / municipal ordinance AI is coming in a very big way. we're going to build the Whole Government Surveillance Machine, and let citizens peer into every interactions of every decision that governs their lives. you can't vote for something you don't understand

@achenfinance Private data (especially in econ/fin) papers will also make this challenging

No we shouldn’t. Despite my AI obsession and months of trying, I found AI auto research scales poorly. I was similarly disappointed by the UCLA Human-x-AI papers (though the competition itself was brilliant).

@iammattduff @barrowjoseph @DenisPeskoff DM’d

@DMilo75 @barrowjoseph @DenisPeskoff very cool project. i've got municode, amlegal, and ecode *for the most part*. the remainder are harder, and critical for my use case because I deal with a lot of rural areas. separately, we seem to have very aligned projects, lmk if you're interested in chatting about it!

New paper: every law in America is technically public. But not really, until now! With @DenisPeskoff at UC Berkeley, we built a corpus of ~every publicly accessibly city and county law, and released a huge chunk of it! 2.2 million laws, you're (probably) covered in it! 🧵

@mycelias @HeySilasReed @barrowjoseph @DenisPeskoff @chiefofscouts @engagicc How far back in time do you go? Also, can you track variance requests/decisions?

@HeySilasReed @DMilo75 @barrowjoseph @DenisPeskoff @chiefofscouts @engagicc UX is consumer focused right now so you can't run queries. i have the beginnings of an MCP that i can formalize if there's interest. capturing voting patterns per elected per item, but only for legistar right now (API access unlike the rest). lifting to the rest possible w/ OCR

@HeySilasReed @barrowjoseph @DenisPeskoff @chiefofscouts The underlying text is public and can be hosted, but it's more challenging to collect and parse PDFs (which nobody continually maintains AFAIK) vs use the structure aggregators (i.e. municode) provide. We're working on the FOIA route, but its costly too.

to @barrowjoseph’s original point, the actual text of municipal code is completely public... and could be assembled the ‘hard way’ with automated FOIA requests, I think? are there any aggregators pulling together an up to date canonical dataset that covers all the data your team has been analyzing, @DMilo75?

@TheStalwart I was excited about LLM tournament scoring at first, but it seems to produce very similar results to asking outright for a score (say 1-10), and costs more/is more complex to understand.

Sorry. Have to brag one more time about getting my first (AFAIK) academic citation. But also I do think the basic approach: Tournament, Normalize, Train is probably a useful one for social scientists looking to turn qualitative, subjective judgment into something quantitative.

@readforyourself @barrowjoseph @DenisPeskoff @DDemocracy_ca @CalPolyIATPP During work on thezoningproject.org, we found some states (especially in the southeast) to charge money for document requests and/or only work with in-state residents. Another hurdle is potential legal ownership from aggregators like Municode.com

This is OUTSTANDING! A true public service. Question: Did you have legal pushback in some jurisdictions. During my time with the @DDemocracy_ca project at @CalPolyIATPP we encountered states that not only claimed their entire state code was copyrighted, but also every word spoken on the floors of their legislature. The lengths some bodies went to to thwart public access, knowledge, scrutiny was incredible. GREAT JOB!!!

@J_Scott_Art @barrowjoseph @DenisPeskoff One window into this is local government meetings, some interesting analysis here: arxiv.org/pdf/2604.21202

@barrowjoseph @DenisPeskoff Do you think thus could be compared against how often these laws are actually applied and enforced? Like could you also cross reference every court case, fine, permit rejection, and arrest? Thatd give a view of not just what the law intends but what it actually does in practice

@bluezenx @barrowjoseph @DenisPeskoff What types of regulations do you think are most important outside of zoning?

@barrowjoseph @DenisPeskoff You should monetize this, what you have here is the bible for doing business in any part of the country.

@HeySilasReed @barrowjoseph @DenisPeskoff @chiefofscouts The aggregators themselves provide this in a sense, and it’s not clear if it’s legal to rehost their data. We’re working towards this though, and for historical codes which we’ve collected. Though, as academics we’re also incentivized to analyze the data extensively first.

@DMilo75 @barrowjoseph @DenisPeskoff @chiefofscouts seems like creating a centralized open source data lake that’s consistently updated could power all of these projects.. and it doesn’t seem like anyone is doing this yet?

@HeySilasReed @barrowjoseph @DenisPeskoff @chiefofscouts We show municode population wise covers a majority of incorporated places and ecode pads out a good chunk of the rest thezoningproject.org

as @chiefofscouts mentioned, there’s an open API that covers almost all of this, no OCR required. a few months back I indexed the 4,281 municipalities that use municode specifically, pulled their entire code down, and then laid the ground work to start tracking changes over time with a weekly diff. on top of that, we can pull meeting minutes from the meetings that created them and track changes in the code back to the legislators that voted for the changes in the first place. you can start to picture the possibilities with this kind of data, especially tracked live and over time. I’d be more than happy to chat with you and the team on this because I agree with all of your foundational notions about how important a project like this is.

@iammattduff @barrowjoseph @DenisPeskoff If you limit yourself to code aggregators like municode and ecode the codes will already be structured and population wise cover most of the US. This allows fast scraping, that we do here thezoningproject.org

@barrowjoseph @DenisPeskoff typically machine readable pdf, but the main problem is I need fresh data. freshness matters especially for energy uses these days, so I can't do one big manual pull. so it's mostly a web scraping problem, and paid services only cover a fraction of the country.

@barrowjoseph @DenisPeskoff Very cool! In related work we answer a set of structured zoning questions on a similar set of municipalities. Dataset/paper available here: thezoningproject.org

@alexolegimas @dwarkesh_sp @pawtrammell Super interesting interview. One thought this inspired: managers likely derive relational utility (i.e. status/power) from managing humans rather than AI. Given the concentration in many industries, managers will likely be able to hold onto this power, delaying AI adoption.

Really enjoyed this episode. Thanks to @dwarkesh_sp and @pawtrammell for the conversation. What I hope that I was able to convey that it is incredibly difficult to make predictions when there is so much uncertainty: there is not just uncertainty around the parameters, but even what model to use in the first place. In my view, the best application of economics to our current moment is not trying to individually forecast scenarios 5 or 10 years out (though aggregate forecasts are useful). There is way too much uncertainty at every level of the exercise. It’s to model important scenarios and work our way backwards: start with a potential scenario that are important to consider and then derive the conditions under which it can arise. This not only allows you to potentially rule out a very intuitive-sounding scenarios because the conditions required are implausible. It also points to data that you need to track which you were not considering before. Eg latent demand for human involvement, substitution between AI and human interaction, task bundling inside jobs, AI bottlenecks, and whether AI looks more like electricity or social media. This is the type of data I’m working to collect, and I know other teams are too. The last point is particularly important. To quote Demis Hassabis, we are potentially at the foothills of the singular. As economists we have the responsibility to guide that transition with both humility and the best information we can gather.

@alz_zyd_ Value menu/daily coupons at fast food (I.e. McDonald’s) fit this category, they’re just a bit hidden on the menu and/or require using the app for coupons.

Relatively poor Americans are not only willing to regularly buy Chipotle, they're willing to regularly Doordash it, roughly doubling the price. America is just so rich that nobody meaningfully cares about paying $6 versus $15 versus $30 for lunch on a regular basis

For a while I've wondered why there's no restaurant food in the US meaningfully cheaper than Chipotle, ~$15 a meal. I think I finally understand: America is so rich that there is just no demand for food much cheaper than this

@OfficialLoganK @rezoundous The value in the Gemini API is unmatched

@rezoundous one of the things I enjoy the most about Google is our willingness to keep shipping and iterating, so pls keep the feedback coming :)

Google's fumble with Gemini and Antigravity needs to be studied.

@toddsaunders I’m in, live in Westfield. I’m a NYU PhD student using AI to study zoning regulations and central bank expectations at scale

I want to get the most AI pilled folks in NJ for a meetup in Westfield. I have no idea how to source or find these people so please X do your thing!

@heeney_luke You can build a Claude code skill that reviews your x bookmarks, downloads ones that are papers, and indexes it to Zotero. You can also add a pdf parser to markdown as part of the process.

Zotero but for all the papers I’ve bookmarked on X, who’s building this?

@alz_zyd_ Learning to game things is an important skill itself. It generally requires building a strong and active network. Beyond that I think 1) peer effects are strong and 2) this pattern reflects risk aversion: variance of outcomes lowest at elite institutions.

Honestly, the huge focus on getting into a good college, PhD program, Goldman Sachs, etc. seems to reflect a deep disbelief in the eventual efficiency of labor markets. Nobody seems to believe any more that if you just develop skills, the skills will pay off eventually

@businessbarista Claude

I want to start a community dedicated to Claude Code. It’s become the gateway drug to coding and experiencing the power of AI for tons of people. This will be a space for people to share killer use cases, agentic workflows, proven prompts, and connect with other CC obsessives. Comment “Claude” if you want to join.