Thomas Monk

and also a VS Code/Cursor/Antigravity/Windsurf extension to run Stata code yourself directly within your IDE, using this MCP in the backend github.com/tmonk/stata-wo…

i'm biased (Claes recommends my Stata Workbench extension, github.com/tmonk/stata-wo…) - but a super useful guide for economists getting into AI within workflows we already use

Proposal: grad students should get a budget for compute

@BeatrizGietner steve machin/sandra mcnally/steve pischke (and many others!) here at the CEP?

ECON FRIENDS: I need help :) My supervisor has collaborated with a lot of people (good for him obviously ❤️), which makes it almost impossible for us to find an *external examiner* in Ireland for my viva Does anyone know an applied micro prof specialized in labour/education that we can invite? Preferably someone in the UK

@caterina_vieira @IIES_Sthlm congrats again - such a great outcome!!

Very happy to join @IIES_Sthlm next academic year as an assistant professor!

@BrendanNyhan completely agree that it's an option. i've moved from the command line to the mcp server. abstraction helps. it allows for multiple sessions, easy data browsing, help files being handled properly, errors being caught etc. all of this is abstracted without requiring more token use

@ttmonk mine just runs it from the command line, is there any benefit to mcp?

built a lightweight mcp server for stata. run commands, check variables, and debug code through claude/codex/antigravity/windsurf without copy-pasting stata output into the chat. ~700 lines of python github.com/tmonk/mcp-stata

@cblatts hi Chris, great to see others getting claude-pilled! just flagging some Stata tools I've made that might be of interest to you/others x.com/ttmonk/status/…

4w ago I was a Claude Code skeptic. I'm not a coder. None of the use cases were relevant. I managed teams & projects, drowning in email & overdue reminders. So I tried creating tools that would help me and... holy crap. Now I'm sharing the tools I built: claudeblattman.com

I am offering mentorship for PhD students in economic history again this year. My PhD benefitted from having a fantastic supervisor, whom I can never repay. He said, "pass it on". I'm not an established academic, but I know what being a PhD candidate is like. 1/2

had a great time here last year - well worth applying!

@YujungHwang3 asking Claude to write tests against expected output, and stating directly that success depends on passing these tests helps hugely with this

@hanjliu i think Stata was inspired by you x.com/Stata/status/2…

going for the distinction in stats related puns

@LSEnews @LSEPublicPolicy credit to @hanjliu

my mpa metrics class is v cute. look at how much fixed effects are improving their lives. @LSEnews @LSEPublicPolicy

@aniketapanjwani the thinking is: is the cli interface designed by StataCorp the best way of interacting with Stata. if not, some light middleware can improve this

@aniketapanjwani i thought the same. but after setting up a workflow almost exactly like yours, the mcp just made more sense. multiple sessions, data browsing, help files being handled properly, etc. all of this is abstracted without requiring more token use by the llm

@alz_zyd_ @cherylwoooo giving it direct access to documentation etc through the mcp directly helps a lot

@alz_zyd_ @cherylwoooo it struggles. which is why i made this github.com/tmonk/mcp-stata

Interesting. 😊I will check this again in a year. The reason I asked was that we keep hearing "AI does 90% of coding" and seeing high SWE-bench scores, but where's the economic growth? One bottleneck: adoption. For example, economists could benefit from Claude Code, but it means switching to Python/R. Tech adoption is slow when it requires changing workflows, not just adding tools. People also mention how Stata is useful for data viewing/quick tabulation commands, etc. I expect there will be more great innovations that would help with the tech adoption, though. e.g., @ttmonk has shared his MCP server that would bridge Stata and Python.

Look at how being diagnosed with cancer increases people's involvement in crime! In the graph, the cancer diagnosis is the vertical dashed grey line. To the left is before the diagnosis; to the right is afterward. At first, leading up to a cancer diagnosis there's no spike in crime. When someone is diagnosed with cancer, their crime declines rapidly. But it doesn't stay down for long. Two years later, people's involvement in crime increases and stays high. Why? The authors provide two reasons: First, economic strains leads individuals to compensate for their loss of legal revenues with illegal earnings. Second, cancer patients face lower expected cost of punishment through a lower survival probability. Wow!

@wdhparker @kevinweil cheers will, having exactly that issue! i’ve tried clearing the history & using the different panes but still get this error across all projects. gutting, hope it can get fixed.

@ttmonk @kevinweil Tom - does your document compile okay, but prompts don’t work? I was having issues where the inline prompt window was giving me only errors. But if you clear the chat history then it works fine. Also there is a separate chat pane on the LHS where you can make larger prompts.

If you're a scientist and/or write in LaTeX, try Prism at prism.openai.com and let us know what you think! Taking feedback and feature requests here and turning them into code daily with Codex.

@cherylwoooo they can use my MCP server and not have to change their workflow!! github.com/tmonk/mcp-stata (even though i am 100% for a full move to python in econ generally)

Do you think >90% of economists will switch to Python/R due to Claude Code, in a year?