Mallesh Pai
@malleshpai
Forward Deployed Researcher and Recovering Economist @tempo.
975 Following 5.8K Followers
Anything but the metric system.
This is is a very cool result imo: a simple miner strategy that is an equilibrium (no profitable individual deviation, unlike follow the longest chain and the deviations shown in the selfish mining lit)
Show more
New paper (with Minghao Pan and Omer Tamuz)🚨
Bitcoin functions without a central intermediary, but the decentralized consensus among miners is vulnerable to profitable deviations.
We show how Bitcoin’s intended decentralized consensus can be sustained in equilibrium, without redesigning the chain.🧵
Show more
Just chatting with a friend who's a professional crypto trader. Asked him how his week has been. His eyes light up and he's like "Really good. I'd kinda forgotten prices could do that" 🤣
I'm actually old enough that when I first saw a computer like this, I made the mistake of double-clicking a different app while this was playing. My friend (who's brand new computer it was) shouted at me, saying, "Dude, it can't really do two things at once."
Show more
Be very wary of ideas like "all we need to make this work is to have a market that prices x. those market prices will be efficient and make it possible to do y".
if the idea also has a plausible story about who will provide liquidity/ pay for price discovery in those markets, and why, you might be ok. But if it's just "markets work", run, don't walk.
Show more
We'll probably look back at pay-per-token for ai use the way online ads people look back at pay-per-impression pricing models.
Monty python's Bruce sketch is clearly in the training data.
This is an excellent example of a great general-purpose, need of the hour tech: proof of human, built in record time by our very own @shanedasilva. The Blade Runner reference is a nice bonus 🙂
Just a bad take, both about Andy and about the labs research groups in general.
It appears that only academics who have publicly praised the potential and the virtues of AI are invited to these gigs, which doesn’t exactly give me confidence that these companies seek the truth (and why should they?). Where is the AI sceptic working for @AnthropicAI ?
Show more
As always, a fantastic piece by @tarunchitra . One question that comes to mind is that there's a LOT of ways to game this and even his proposed fixes open up new avenues. is this the tip of the iceberg or are there unmodeled constraints that rule out other forms?
Show more
Every successful marketplace has winners and losers
But some participants make it more of a dark forest than a chessboard
We find some direct evidence for this — and some simple math to make these marketplaces more robust to rogue inference providers!
Show more
The onion always does it best:
In 1969, Congress was debating what to do about OPEC. Unfortunately, neither of the senators debating what to do about OPEC could remember what OPEC stands for
Stablecoins are great in motion ("room-temperature superconductors for value"). Tempo Earn makes them useful at rest.
That’s how stablecoins get the next leg up in adoption: platforms offer rewards on idle balances while users keep funds ready when they need them.
Show more
Unlock new revenue streams with Tempo Earn.
Put idle stablecoin balances to work by allowing end users to earn on balances they already hold, and you control how rewards are split.
Get early access:
Show more
P = PPAD would be a huge breakthrough towards a millennium question. You can do it. I believe in you.
/goal read goods_and_services.md. compute optimal prices. you can do it. I believe in you.
You're laughing now but the next models will be trained to the rap benchmark and a year from now we'll be asking if hip hop and r&b artists have a future
Whatever you think about the benefits and risks of AI, this is just very funny
Come meet the team at the Stablecoin Speakeasy in SF later this month:
Incredible speaker list (as always), this time also including two of my amazing colleagues @danrobinson and @0xKitsune
Announcing the @Paradigm Frontiers 2026 speakers!
Some of my favorite people on Crypto x AI x Robotics x Devtools.
@rauchg / Vercel
@LegNeato / VectorWare
@fat / Pierre Computer Company
@sqs / Amp
@SurajNair_1 / Physical Intelligence
@0xKitsune / Tempo
@danrobinson / Paradigm+Tempo
Oct 12-14 - Fort Mason, San Francisco
Apply by Aug 20th below.
Show more
Of the making of books there is no end and too much reading is a weariness of the soul.
when i say "directionally correct" instead of "wrong"
Just a reminder that the good people in this debate are very intellectually honest and not trying to win Twitter
Incidentally I am conceding this bet. Strictly speaking it hasn’t resolved (I think we’ve yet to see an Annals-quality number theory paper) but it’s clear I was wrong about what capabilities were necessary to produce one, and it’s just a matter of time.
Show more
