Aurash Vatan

Economics of AI research @mercor. Formerly @DEShawGroup @thecrimson

[Aurash Vatan @logorrhythmic]

90% of Americans’ working hours are spent on tasks not covered well by AI benchmarks.

That’s our best guess from a new project mapping frontier benchmarks onto the US economy.

Preparing for a post-AGI economy requires predicting the future. Current benchmarks makes it very difficult even to measure the present.

[Aurash Vatan @logorrhythmic]

@PunditPool Designed for a very specific kind of nerd

[Pundit Pool @PunditPool]

Do you think you can call the elections?

Prove it 👀

Pundit Pool is a 2026 prediction game:
🦅 Basic: 5 calls per chamber
🔥 Extreme: all 77 battlegrounds
📈 Forecast: every race + margins

Make a group with your friends!
https://t.co/IHzSdaBvTt

[Thomas Hikaru Clark @thomashikaru]

If you've noticed a distinct dialect emerging in frontier AI models and struggled to understand what they're talking about, you're not crazy!

We turned to psycholinguistics to investigate what makes the writing style so weird, and why this might be happening

🚨New preprint! 🧵

[aden barton @aden_barton]

Looking forward to joining MTS later today to discuss the Mercor Econ team's recent work!

[Aurash Vatan @logorrhythmic]

@cattttang @neilmovva @substratemag Where are the short-form vertical-vid outtakes @cattttang for those with shorter attention spans

[Aurash Vatan @logorrhythmic]

@aden_barton Let's go!

[Aurash Vatan @logorrhythmic]

If a true nerd, sparkle will be unobservable due to eye contact aversion. This is what makes VC investing so difficult.

[Paul Graham @paulg]

My favorite kind of person is the curious nerd with a sparkle in his eye.

[Aurash Vatan @logorrhythmic]

@SamiBelhareth We actually quantify (b)! Would be very interesting to look at regulatory exposure in the next iteration of this index. https://t.co/rnMBLQA6Bl

[Aurash Vatan @logorrhythmic]

Benchmarks prioritize high wage-bill, AI-exposed jobs & tasks. This makes sense.

But the rest of the economy becomes more relevant as models diffuse. We want to know what AI means for paralegals and medical assistants too, not just doctors and lawyers.

[Aurash Vatan @logorrhythmic]

@paulg If a true nerd, sparkle will be unobservable due to eye contact aversion. This is what makes VC investing so difficult.

[Aurash Vatan @logorrhythmic]

@rocketalignment https://t.co/9OudljcJjg

[Aurash Vatan @logorrhythmic]

Many benefits of AGI won’t show up in GDP, coming instead in the form of higher-quality tweets:

[Aurash Vatan @logorrhythmic]

@willdepue Good example of instrumental convergence

[Aurash Vatan @logorrhythmic]

@JasminLovesAI It’s a good q. If we really think these benchmarks are public goods, that suggests optimal coverage requires funding from gov’ts or non-profits — I think this solution is underdiscussed.

[Aurash Vatan @logorrhythmic]

90% of Americans’ working hours are spent on tasks not covered well by AI benchmarks.

That’s our best guess from a new project mapping frontier benchmarks onto the US economy.

Preparing for a post-AGI economy requires predicting the future. Current benchmarks makes it very difficult even to measure the present.

[Aurash Vatan @logorrhythmic]

Nordic data would be better for measuring IQ, but presumably wage effects would be slower there, and AI adoption is decently lower there too per @AnthropicAI Economic Index

[Aurash Vatan @logorrhythmic]

@OppInsights should check if the IQ wage premium has changed post-ChatGPT using their linked SAT/IRS data. @LichtingerGuy & @SeyedMH98 suggests impact should be U-shaped, but admin data would be so good here.

[Aurash Vatan @logorrhythmic]

Full blog here, plus an interactive Benchmark Map you can click through. This is joint work with Isaac Robinson and @aden_barton , with invaluable support from the rest of the Mercor research team. https://t.co/MPKYDEvjmF

[Aurash Vatan @logorrhythmic]

Of course, there are limitations to this work. We base our analysis on an initial set of 29 economically relevant benchmarks. We sought out a broad sample of high-quality, unsaturated benchmarks, but selection might still skew our results in various ways, and any incomplete sample necessarily biases down coverage estimates. We will continue to expand our benchmark list over time, updating these results as we do so.

[Aurash Vatan @logorrhythmic]

Just got my 2025 tax refund, fortunately this won’t be changing my consumption at all @NobelPrize @riksbanken