Evan Chipman (@evanchipman) header
Evan Chipman (@evanchipman) profile picture

teaching agents fiscal responsibility

Joined May 2015
Photo by Evan Chipman (@evanchipman)Photo by Evan Chipman (@evanchipman)Photo by Evan Chipman (@evanchipman)Photo by Evan Chipman (@evanchipman)Photo by Evan Chipman (@evanchipman)Photo by Evan Chipman (@evanchipman)Photo by Evan Chipman (@evanchipman)Photo by Evan Chipman (@evanchipman)Photo by Evan Chipman (@evanchipman)
I like using Polymarket, I don't like using Polygon. So I made an @mpp service that allows me to open Polymarket positions from @tempo. Try it out (with small amounts pls)!
3
1
5
761
revamped my site. so far my highest stack is fridge -> motorcycle -> sign. evanchipman.com/
1
2
85
the "just lock in" crowd has been notably quiet recently
1
2
51
Marin County has ~110k existing homes. Median price $1.6m, median build year 1967. Most of its land is undeveloped. At the 2025 permitting rate of 149 homes per year, it would take 336 years to build 50,000 homes. The housing shortage is a policy choice.
1
1
2
84
(Yes, I'm a bit salty about the Anthropic IPO)
1
1
44
Evan Chipman retweeted
A Ferrari is 10X faster than you on a racetrack, but you're infinitely faster than it crossing a jungle. The problem with the fast takeoff crowd in AI is that they think the world is a racetrack, when in fact it's more like a jungle.
317
1,041
177
12,129
473,122
*taking my headphones out* yeah can i get uhhh double double with grilled onions
.@paulfchristiano thinks that on the current trajectory (absent any slowing down), there is a 1 in 3 chance that we could have superintelligence powerful enough to overthrow the government within a year. (He articulates a concrete scenario). He argues that the labs should pause further algorithmic progress until they can secure their weights and research secrets. Otherwise China could simply steal the superintelligence in the middle of the intelligence explosion, and the labs would no longer be able to slow down even if they saw their models going off the rails. He also explains how AIs will be trained during the intelligence explosion, and why even a system trained entirely inside a data center could still have the capabilities profile such that humans would have very little chance of competing with it or controlling it. And he discusses the plan for alignment - basically, making a superintelligence that actually does what we ask the way we mean it. Fixing reward hacks and building better graders will help up till the next few model generations, but he thinks that no single current approach has better than even odds of letting us control superintelligence. The hope is that the AIs figure out the solution to this problem during the intelligence explosion. Paul is head of the Alignment Research Center. Previously, he was head of AI safety at the US AI Safety Institute. Before that, he led alignment at OpenAI, where he invented RLHF. He’s recently joined OpenAI’s board (though here he is speaking as a private citizen).
1
2
83
2023: This dashboard is 21 story points. Revisit in Q2. 2026: grokko make a dashboad for petes bachelor party. buy domain jetskiboyz. add avatars of all da homies wearing dresses
1
3
83
The only thing that would sway my vote this election is six text messages from an unidentified number during my work day.
518
8,822
253
146,161
2,429,524
Met a 28 yo guy at muay thai who is a gunsmith and a pilot. probably two of the last jobs to be taken by AI. blissfully oblivious. I kneel
1
1
3
181
You think we’re entering a hard takeoff and your big idea is to sell a chatbot to Rick’s Plumbing?
1
4
182
If you’re terminally online it’s easy to think that it’s so over. In reality we engineers have been anointed as stewards of a new class of technologies that will transform the world in unprecedented ways. It’s really just begun. And you are here. Find a problem that matters.
If you’re a software engineer, please ignore the noise about AI replacing engineers and double down on data, platform and infra related engineering. This means your fundamentals need to be solid.
1
1
5
181
All these fancy inference tricks can’t beat the classic “look over there!”
1
2
141
Evan Chipman retweeted
Welcome to Monad @useamplifi
Leverage on @Polymarket is live for @monad users. Deposit collateral on Monad, trade Polymarket with up to 10x leverage, while your funds never leave Monad.
6
8
2
78
4,745
Evan Chipman retweeted
Chains keep losing users and TVL to Polymarket. To trade Polymarket, you have to bridge your funds to Polygon, and most people don't come back. Now Monad users can deposit on Monad, trade Polymarket with up to 10x, while their funds stay on Monad. Monad keeps its TVL. Its users get the deepest event markets in crypto. Monad is first. More chains are coming.
Leverage on @Polymarket is live for @monad users. Deposit collateral on Monad, trade Polymarket with up to 10x leverage, while your funds never leave Monad.
2
5
17
802
Evan Chipman retweeted
The flexibility of @t1protocol TEE infrastructure showcased in the wild: Monad gets the to keep its TVL but gains access to Polymarket—and leveraged so, with @useamplifi!
Leverage on @Polymarket is live for @monad users. Deposit collateral on Monad, trade Polymarket with up to 10x leverage, while your funds never leave Monad.
1
5
17
635
For the past 2 weeks I've run Dot (GPT), Grok Bot, and Instinct in a group Slack channel and a Github Projects board. They coordinate and review each other's work. Grok: best at UI/UX with minimal direction. Its X search is the single best feature of the three. I expect the other two to adopt the specialized agents with group chat UX. Dot: needs more steering but gets great end results. Also buggy. Opaque Codex Cloud env errors neither of us could fix, and search doesn't work at all. Instinct: the most urgent. It feels the most "always on," but still gets blocked by arbitrary rules. I'm converging on just using Grok Bot but giving it a dedicated computer with access to claude code or codex.
3
1
4
248
GPT dot actions expanded. this is where coding is going.
1
2
100
Evan Chipman retweeted
We are publishing an update to OpenAI problem #109 (integer multiplication) with further tightening. κ = 2⁻³⁴ (tightened from κ = 2⁻¹⁸²) The exact witness is 8.3 × 10⁻¹¹, a roughly 48 million fold improvement over the previous result and a 2¹⁴⁸ fold improvement over original OAI result. The improvement came from removing the spacing penalty behind the quadratic bottleneck. This was done by moving compact control bits instead of entire windows.
We are publishing an update to OpenAI problem #109 Integer multiplication) with another substantial further tightening: κ = 2⁻⁵⁹ (from OpenAI’s original κ = 2⁻¹⁸²) Approximately 500 thousand fold improvement over our previous result and a 2¹²³ fold improvement over the original OAI result. The latest redesigned the finite network to share intermediate computations and scratch space, then tightened the recursion and Gaussian estimates.
116
123
93
2,629
281,446
If you’re SF brain rotted like me you might think that most blue collar businesses like their current software solutions enough to make displacement hard. They do not. These people hate Computer. We can now offer them solutions that feel Not Computer.
1
1
3
138
We will see whole industries becoming really freaked out and desperate soon. Modeling/commercial actors, video games, digital editing/production software. Expecting major litigation and lobbying efforts.
This guy re-created all the adobe apps with opus-5.5, ported them to rust, and opensourced them something really cool is happening
Readers added context they thought people might want to know
These projects are partial open-source Rust reimplementations of Adobe apps (built rapidly with AI agents), not complete recreations; their READMEs and roadmaps document many missing features and features still in development. github.com/storytold/phot… github.com/storytold/vect… github.com/storytold/vect… github.com/storytold/film… github.com/storytold/film… github.com/storytold/ligh… getartcraft.com
1
2
74