Ecomatic

AI × Economics | Econ & Business student

[Ecomatic @GetEcomatic]

I'm a business & economics student, deeply interested in what AI is doing to the economy.

This account is where I document that.

What you'll get here:
→ Real analysis of AI's economic impact — jobs, markets, business models

→ What I'm learning as I go through my degree, in real time

→ The right calls and the wrong ones.

Follow along! Watch it build.

[Ecomatic @GetEcomatic]

A team of mathematicians has university funding locked in through 2029 to do a job Claude just finished in 11 days.

Nobody's saying it out loud, but that funding just became very hard to justify.

This isn't a story about math. It's a preview of what happens to every field where the work is "verify what's already true."

[Ecomatic @GetEcomatic]

@Kalshi Somewhere at Imperial College a mathematician with funding through 2029 is refreshing this tweet and quietly recalculating their timeline.

[Ecomatic @GetEcomatic]

UBI isn't compensation. It's a subscription fee for staying quiet while a handful of people own the machine that made your job obsolete.

They'll dress it up as progress. It's actually the smallest possible payout to avoid a real conversation about who owns what.

You're not being saved. You're being managed.

[Ecomatic @GetEcomatic]

@Contrarian03 The mechanism is real, big IPOs do pull liquidity toward equity. But crypto and equity aren't one shared pool, most crypto money never touches a brokerage account at all. If this makes sense.

[Ecomatic @GetEcomatic]

Building Ecomatic on one bet: most people covering AI either hype it or dunk on it. Almost nobody actually tracks where the money moves.

If you're doing real work in AI or economics, reply with what you're building or writing. I follow back people actually doing the thing, not just posting takes about it.

[Ecomatic @GetEcomatic]

Circular AI money is the story nobody's pricing in correctly.

Nvidia invests in OpenAI. OpenAI buys compute from Oracle. Oracle buys chips from Nvidia.

Three companies trading the same dollar in a loop and calling it revenue growth.

This is not new, it's vendor financing with better branding. Works fine until one link in the circle needs actual outside cash.

[Ecomatic @GetEcomatic]

@aniket_ai That 15x number is the real story, not reliability. Fewer engineers per unit of output eventually shows up as headcount pressure, even without full autonomy.

[Ecomatic @GetEcomatic]

No, coding agents aren't replacing engineers yet.

Every company ships the demo, never the failure rate. If it actually worked without a human checking the output, nobody would still be paying for that human.

[Ecomatic @GetEcomatic]

@FpeSre Exactly, and that is the quiet part. Every lab benefits from talking about responsible rollout while the gap itself is the actual competitive moat. Slower public releases are not just caution, they are a feature nobody is pricing in.

[Ecomatic @GetEcomatic]

Every AI lab has two versions of their model. The one employees use internally, and the one that ships months later once it's safe and boring enough for the public.

OpenAI just admitted this out loud. Astra was good enough to pull their roadmap forward 6 months before anyone outside the building touched it.

The real AI race isn't who has the best public model. It's who has the biggest gap between what they use and what they sell you.

[Ecomatic @GetEcomatic]

@DanDr1s Four model names plus an image update in one DevDay isn't "stacked," it's throwing everything out before Gemini or Anthropic eat the news cycle. Volume isn't the same as importance.

[Ecomatic @GetEcomatic]

Germany is building data centers at an unprecedented scale while burning political capital on energy supply fights nobody in Brussels has actually solved yet.

Meanwhile the actual AI race is being run on capital efficiency, not electricity headlines. Whoever trains the cheapest model per unit of capability wins the next round, not whoever announces the biggest concrete pour.

Europe keeps talking about becoming the first AI continent. The compute, the capital, and the talent pipeline all say otherwise. Building infrastructure after the leaders already have theirs isn't leading, it's catching up expensively.

[Ecomatic @GetEcomatic]

$344 a month at 65 plus Social Security is not retirement, that is a very committed subscription to being anxious. Curious how that number looks once you adjust for the 28% of accounts holding under $10k, since even $103k median for that age bracket is being propped up by people who are doing fine dragging up everyone else's picture again.

[Ecomatic @GetEcomatic]

@andymewborn Only works if you can actually tell signal from noise in one data point, which you usually can't.

[Ecomatic @GetEcomatic]

I think if you build one AI tool that suddenly takes off, gets shared everywhere, makes you real money in a week, the best thing you can do is not touch it for a few days.

It is easy to think you cracked some code after one project blows up, and a lot of people treat that as proof of skill instead of proof of timing, then bet everything on repeating it immediately.

So yeah, maybe you picked the right niche. Maybe the timing lined up. Maybe you got lucky on distribution and the algorithm just decided to like you that week.

Take the win. Actually use it. Then step back and ask what part was you and what part was luck, before you quit your job or dump your savings into scaling it.

Because the dangerous part is believing you have found a formula. You start skipping the boring validation work because you feel like you already know the answer. That is usually when the next project quietly flops and you cannot figure out why.

One win does not mean you understand the game. It means the game was kind to you once. Take the data, stay humble, and go build the next thing with a clear head instead of an ego.

[Ecomatic @GetEcomatic]

@ipanditshashi Can you explain the win in one sentence without describing luck? If yes, skill. If it's just what happened, luck.

[Ecomatic @GetEcomatic]

@andymewborn They are moving fast.

[Ecomatic @GetEcomatic]

Anthropic and OpenAI are both shipping coding agents that finish tickets end to end, not just autocomplete.

That fits directly with where $MSFT and $GOOGL are pushing Copilot and Gemini pricing, because once an agent can close the ticket itself, the seat stops being the product and the completed task becomes the product.

[Ecomatic @GetEcomatic]

@Contrarian03 Honestly the timing on any single stock would be impossible to call. The interesting trade isn't the day of the announcement, it's watching seat count growth slow over the next few earnings calls while usage keeps climbing.

[Ecomatic @GetEcomatic]

Trivial for Sam Altman with unlimited compute and the model's own team on call is not trivial for the actual indie developers this is supposedly about to disrupt. The gap between a demo that works for the guy who owns the lab and a tool that ships reliably for everyone else is where most of these AI game claims quietly die.

[Ecomatic @GetEcomatic]

@elonmusk 600 landings and yet somehow my food delivery still shows up at the wrong door. Curious what the actual failure rate looks like now, since nobody posts a highlight reel of the ones that missed.