Nostr notes by dan

this was very good and articulated a lot i've been feeling ...

2026-10-21T23:10:25Z

this was very good and articulated a lot i've been feeling lately.

in formalization, this effect is even more noticeable: decoherence is punished by the agent being stuck in the maze for days when otherwise it would've made progress in 30 minutes

but i think this applies broadly to software

rip to the real one ...

2026-10-19T01:13:44Z

rip to the real one

https://eurosky.social/xrpc/com.atproto.sync.getBlob?did=did:plc:fpruhuo22xkm5o7ttr2ktxdo&cid=bafkreieddcsshl4g7xjno4e7utopfxol6g5vwd7wz6juagymo4bc6a7rny

i’ve been doing some research on using Lean with AI to prove ...

2026-10-04T16:11:13Z

i’ve been doing some research on using Lean with AI to prove software properties (in my case, pieces of React). i’ll try to share lessons if it gets to a useful state. the main thing i’m learning is that, with this setup, proofs are *grown*. you can’t “just build them”

fun read Our Eukaryotic Moment

2026-10-04T14:29:26Z

fun read

Our Eukaryotic Moment

Nostr event nevent1qqsdef2spg0x28k8affsgrjxr7dp4has2dmtz03phq4g9sztf407geszyrat22ls2dd5tp0gy4syvwzxxteakz35h3we7jdg7qmclc47pp2zjd3wjc0

2026-10-04T12:04:22Z

To grieve, or not to grieve?

Nostr event nevent1qqsrd44s7x9wlv45kq0w4q35w0typmlw8fztlt5ag24h86qysm9nl9gzyrat22ls2dd5tp0gy4syvwzxxteakz35h3we7jdg7qmclc47pp2zjc2unkm

2026-10-04T01:25:00Z

been unironically upping my p doom over the last few months

in my experience, AI is excelling at most technical work except ...

2026-10-02T11:50:45Z

in my experience, AI is excelling at most technical work except for ”what is the right next problem to work on”. for now, humans still have a noticeable advantage there

hey this album slaps Pure Pleasure For The Piano - ...

2026-09-30T21:13:09Z

hey this album slaps

Pure Pleasure For The Piano - ...

Nostr event nevent1qqsr5eug5rdzn2qzag97fzsj52jeta08q74r4atg09z2p4gn786ylrgzyrat22ls2dd5tp0gy4syvwzxxteakz35h3we7jdg7qmclc47pp2zjupzdc0

2026-09-28T20:00:29Z

astra may be my favorite model now. it's a very weird one

this seems like it should be fun The World as a Hologram

2026-09-25T19:51:39Z

this seems like it should be fun

The World as a Hologram

Nostr event nevent1qqstycqcm5eews0n9dtlw6z9qjmxj6zfvr20uf6y38dnz6j299pepsqzyrat22ls2dd5tp0gy4syvwzxxteakz35h3we7jdg7qmclc47pp2zjvskuqj

2026-09-24T11:56:10Z

guys

https://eurosky.social/xrpc/com.atproto.sync.getBlob?did=did:plc:fpruhuo22xkm5o7ttr2ktxdo&cid=bafkreib5vov3wvyfgnglvbbit46c3tf2vtwm43y2nbmx7pjrakld76a5yy

ok i take it back Astra is good how to work with it: 1. get it to ...

2026-09-23T13:45:47Z

ok i take it back Astra is good

how to work with it:

1. get it to deeply understand what needs to be built. not “plan“ but like get it to be a domain nerd

2. only THEN, give it a usable past sloppy attempt and ask for excellence

3. ask it to use Sols for coding so it stays on strategy and taste

Nostr event nevent1qqs0vxaszy7mrj4p4kxjxu8ve7jt37vdrwnkd2d3vj3dnxff8h5wafczyrat22ls2dd5tp0gy4syvwzxxteakz35h3we7jdg7qmclc47pp2zjcqh9va

2026-09-23T01:21:53Z

Minna-no-kimochi (みんなのきもち) | B...

Nostr event nevent1qqsq4mrquqmkxu56v50qep34tv3r8syv6490k56h05p6q2aca8l0vdqzyrat22ls2dd5tp0gy4syvwzxxteakz35h3we7jdg7qmclc47pp2zj0wq8l9

2026-09-22T20:21:34Z

POiSON GiRL FRiEND | Boiler Ro...

thanks to Dr. Samuel Allen Alexander for making a video about my ...

2026-09-20T12:10:41Z

thanks to Dr. Samuel Allen Alexander for making a video about my post! www.youtube.com/watch?v=CFkr...

Conway's Conjecture AI-proved,...

nostr:nevent1qy2hwumn8ghj7un9d3shjtnyd968gmewwp6kyqpqkypzsmdutm33w8gt4l3n87lld5c0c335rfu8hg0ghdyxd060fumsf9uttn

astra is very hit and miss. can get complex stuff done right and ...

2026-09-05T23:04:01Z

astra is very hit and miss. can get complex stuff done right and then goes wildly offtrack

love this solo ...

2026-09-05T10:10:42Z

love this solo

https://eurosky.social/xrpc/com.atproto.sync.getBlob?did=did:plc:fpruhuo22xkm5o7ttr2ktxdo&cid=bafkreifz2y35vgcremhqd7iyyqhnm4oxd2wbeg7rtayemeplgo5znpyvrq

preparing to publish my vibecoded math lean proof to github and i ...

2026-08-30T23:01:35Z

preparing to publish my vibecoded math lean proof to github and i kinda feel nervous cause what if it's all wrong

found the magic prompt that fixes all my Lean. "walk the ...

2026-08-30T12:17:00Z

found the magic prompt that fixes all my Lean. "walk the abstraction layers one by one from the foundations all the way to the top, and at every point see if the proof matches how you would explain 'why' to a human. if not, assert underlying facts at the lower level and compose or lift them"

Nostr event nevent1qqsvtd0rasqk6qj85npssagyksnyqd0zrvt3xa9x6u956hf0ycwlq5czyrat22ls2dd5tp0gy4syvwzxxteakz35h3we7jdg7qmclc47pp2zj6t77px

2026-08-29T16:47:18Z

learning a ton about proof engineering in practice

fantastic tool for generating Lean proof maps ...

2026-08-29T15:33:35Z

fantastic tool for generating Lean proof maps ("blueprints") directly from the Lean code github.com/hanwenzhu/Le...

highly recommend for helping LLM make sense of the broader structure of the proof and finding simplifications

GitHub - hanwenzhu/LeanArchite...

hear me out surreal numbers — сверхчисла omnific ...

2026-08-28T17:03:15Z

hear me out

surreal numbers — сверхчисла
omnific integers — всецелые

1) go back to the initial commit 2) tell sol to port (parts of) ...

2026-08-25T00:44:26Z

1) go back to the initial commit
2) tell sol to port (parts of) the system to lean, converting code into lean definitions and tests to lean proofs
3) repeat for each commit where the system significantly evolved; update definitions and proofs to account for new constraints expressed as theorems

i think the main thing i learned with my llm math detour (and now ...

2026-08-25T00:13:27Z

i think the main thing i learned with my llm math detour (and now see with speculative llm projects in general) is that telling llm to burn it all down and start from scratch when you learn an important lesson (or ask it what it wants to salvage) brings as much, if not more, value than construction

there’s probably a fun blog post to write about the experience. ...

2026-08-23T20:10:22Z

In reply to nevent1q…rsd0
_________________________

there’s probably a fun blog post to write about the experience. anything you want to know?

the most viable route is probably to collaborate with a living ...

2026-08-23T20:06:13Z

In reply to nevent1q…wp7g
_________________________

the most viable route is probably to collaborate with a living mathematician who can take responsibility for the paper form of the result. i believe the Lean statement is expressed right and the project is set up right, and Lean kernel certifies the machine version of the proof.

now the question (pending further verification) is what to do ...

2026-08-23T20:03:27Z

In reply to nevent1q…65rv
_________________________

now the question (pending further verification) is what to do with the result. in mathematics when you publish, you take responsibility for what is written. the only thing i can truly take responsibility for is the Lean *proposition* because i intentionally expressed it using Mathlib only.

the proof itself is fully verified with Lean 4 with no extra ...

2026-08-23T19:56:34Z

In reply to nevent1q…rcrs
_________________________

the proof itself is fully verified with Lean 4 with no extra axioms so that further bumps my confidence. there’s still a chance something‘s wrong with my statement but it seems safe enough to assume it survives. as a part of this, my agent formalized a bunch of results from underlying literature

just got a tentative confirmation from a mathematician that i was ...

2026-08-23T19:52:08Z

just got a tentative confirmation from a mathematician that i was able to get a genuinely novel and interesting mathematical result in a niche area (generalized power series and omnific integers Oz). the actual proof is 100% AI but it took me shepherding it for a month (it did not one-shot at all)

even if you're not literally porting the code i think ...

2026-08-22T23:17:56Z

In reply to nevent1q…4sch
_________________________

even if you're not literally porting the code i think it's good to have a written-down model of each cohesive part of the system. it's just design discipline. but you can also try things "in spec" and see how they affect everything else. like constraints on API design. so that excites me

you can also pin it with ports of integration tests that look ...

2026-08-22T23:16:58Z

In reply to nevent1q…rqmv
_________________________

you can also pin it with ports of integration tests that look close to original tests but are actually a Lean DSL that is also just typechecking theorems under the hood. so you get more assurance you're pinning the right objects. and if your code is FP-lite or C-lite, it translates pretty closely

so this lets you have a "proof-carrying" specification ...

2026-08-22T23:15:05Z

In reply to nevent1q…8ddd
_________________________

so this lets you have a "proof-carrying" specification that lets you look at the system from different levels of abstraction and lets you assert what you believe about each layer and tells you when a change you make at one layer breaks the beliefs of another layer. and the AI can write it

this is not proving things about JS code, but JS code is also not ...

2026-08-22T23:12:52Z

In reply to nevent1q…ejen
_________________________

this is not proving things about JS code, but JS code is also not "React". the actual "React" is the intent behind it and the design spec. so my point is more that Lean is mature enough that you can use it for spec'ing software

the gamechanger is of course AI. it's great at Lean

if you have "theories" at different level of detail, you ...

2026-08-22T23:11:11Z

In reply to nevent1q…f8p7
_________________________

if you have "theories" at different level of detail, you can relate them. this kind of thing happens in mathematics too so it's unsurprising that Lean can handle that.

this lets you prove cross-cutting properties about them. for React, things like "a tree never tears" or "updates don't get stuck"

this is an experiment and it might not work out but it seems like ...

2026-08-22T23:09:22Z

In reply to nevent1q…kkt3
_________________________

this is an experiment and it might not work out but it seems like you can distill the "mental model" or "theory" behind different pieces of react design into Lean data structures and algorithms. you can do this from a very detailed layer (e.g. mirroring implementation) to coarser "theories"

Nostr event nevent1qqsy3udpf3mts09u8uxjk2t65hc4mjf6vrtt0t9469wx7g4e449qs6qzyrat22ls2dd5tp0gy4syvwzxxteakz35h3we7jdg7qmclc47pp2zjp2ls49

2026-08-22T23:08:12Z

i'm formalizing React in Lean (???)

tfw you're a non-technical pm on a software project ...

2026-08-14T00:36:56Z

tfw you're a non-technical pm on a software project

https://eurosky.social/xrpc/com.atproto.sync.getBlob?did=did:plc:fpruhuo22xkm5o7ttr2ktxdo&cid=bafkreigflbsueewyjcvpqait54ytnhe5mpqtfey7beigku4h6kx4edqm4i

either likers have to be present in the appview response for the ...

2026-08-10T13:57:29Z

either likers have to be present in the appview response for the feed itself so they're instantly available on click, or you should fold the bar below into the glimmer, or don't show this at all

@samuel.fm @darrin.bsky.team@bsky.brid.gy when the design ...

2026-08-10T13:55:38Z

@samuel.fm @darrin.bsky.team@bsky.brid.gy when the design requires a layout shift during page load, hand it back to the designer!

https://eurosky.social/xrpc/com.atproto.sync.getBlob?did=did:plc:fpruhuo22xkm5o7ttr2ktxdo&cid=bafkreihoko6favdhwqbin7oyqa7ubcsmqa7s55m4q5o6cjenanbf3znfhq

in either case the question has no relation to javascript as a ...

2026-08-05T21:51:46Z

in either case the question has no relation to javascript as a language. if you could run php on the client, you'd have the same tradeoff of "now i have a universal system and get to decide what to ship as data and what to ship as code"

both server-first and client-first approaches have strength and ...

2026-08-05T21:50:27Z

both server-first and client-first approaches have strength and weaknesses. obviously you won't see churn in each paradigm in isolation because it's simply given up on doing "the other side" anywhere close to well

conversely, you should expect some churn in paradigms that actually bridge them

5% left but fifth time’s the charm amirite #nevent1q…9t7n

2026-08-03T21:11:34Z

5% left but fifth time’s the charm amirite

quoting
nevent1q…9t7n
if it works out i'll have a really fun post to write

love this stuff so much www.infinitelymore.xyz/p/omnific-in... ...

2026-08-03T21:09:20Z

love this stuff so much www.infinitelymore.xyz/p/omnific-in...

https://eurosky.social/xrpc/com.atproto.sync.getBlob?did=did:plc:fpruhuo22xkm5o7ttr2ktxdo&cid=bafkreibeg3t5scymum3gpxdo27ymwzlyp64wam7z2f6skwoxjhexlifqje

Nostr event nevent1qqs8549cjx6yhf83ht74sejr2280p94ljcqklf0cees34j4afd0agvczyrat22ls2dd5tp0gy4syvwzxxteakz35h3we7jdg7qmclc47pp2zj43fysv

2026-08-03T01:52:41Z

if it works out i'll have a really fun post to write

Nostr event nevent1qqs86zutgv7ec058vnj0l9fqtjjas96l58thdmlgh0xdc66xg0n5k3szyrat22ls2dd5tp0gy4syvwzxxteakz35h3we7jdg7qmclc47pp2zjv66ezd

2026-08-03T01:49:34Z

doing a weird llm math experiment

Tau Ceti model is wild. people write proposals and criteria for ...

2026-08-01T23:41:15Z

Tau Ceti model is wild. people write proposals and criteria for how to review stuff, everything else is handled by AI workers

taucetiproject.github.io/TauCeti/

https://eurosky.social/xrpc/com.atproto.sync.getBlob?did=did:plc:fpruhuo22xkm5o7ttr2ktxdo&cid=bafkreidjs2atpqfq2evx2sqk2zpmloj5ywudhtszny4yd3jwpcgmtsx5qm

stuff like "can you enumerate different objects that are ...

2026-08-01T23:32:33Z

stuff like "can you enumerate different objects that are implicated in this, and questions we'd like to ask about them? maybe some things we know or some things we don't know"

i want some cool prompts for remixing knowledge or looking at it ...

2026-08-01T23:32:10Z

i want some cool prompts for remixing knowledge or looking at it from a different angle

Nostr event nevent1qqsdk82x4ajs4jtydeuecfwvjf2t002z030lechxkahpl5qznmhlmlgzyrat22ls2dd5tp0gy4syvwzxxteakz35h3we7jdg7qmclc47pp2zjr9h075

2026-08-01T22:57:37Z

oh shit Sol usage limit is reset! we're so back

man i can finally read papers now. what a time to be alive ...

2026-08-01T20:46:04Z

man i can finally read papers now. what a time to be alive

https://eurosky.social/xrpc/com.atproto.sync.getBlob?did=did:plc:fpruhuo22xkm5o7ttr2ktxdo&cid=bafkreicxnep5n4ygc5xmmdvfgiycjkyuek2phsiefm7ruxftiolkcn25xa

"one weird trick" to get a good agent: on start, tell it ...

2026-08-01T20:41:25Z

"one weird trick" to get a good agent: on start, tell it to look for mistakes in the paper's proofs. this forces it to actually "think through" each clause so by the time it wakes up, its context carries a detailed understanding of the arguments (even if the arguments are correct)

in the end i got a super long tex file (which survives ...

2026-08-01T16:10:48Z

in the end i got a super long tex file (which survives adversarial verification by same models) and some Lean stuff that i'm still grinding. they also found some mistakes in published papers

my current plan is to extract those mistakes + prove them with Lean and contact the authors to see if legit

then i ran out of Sol credits and tried to move this workflow to ...

2026-08-01T16:02:12Z

then i ran out of Sol credits and tried to move this workflow to Claude Code

this was a mixed bag. the Lean track remained pretty good although i have to kick it every 10 minutes because the stupid thing keeps pausing on partial progress

the math tracks were a mess. mistake after mistake

Codex is WAY better than Claude Code. in particular the ...

2026-08-01T16:00:18Z

Codex is WAY better than Claude Code. in particular the “Goals” abstraction in Codex was amazing

i had one session (“PM”) whose goal was to advance the conjecture, managing a few other sessions (exploratory, math, Lean verification flow) and setting goals for them

they mostly stayed on track

Sol seems WAY better at math it found a bunch of seemingly new ...

2026-08-01T15:57:35Z

Sol seems WAY better at math

it found a bunch of seemingly new things that survive adversarial verification by both. they’re not huge results but seem new and not in the literature

Fable by contrast seems to either descend into insta-giving up or kept coming up with ultraslop nonsense

ran out of max credits on both Sol and Fable trying to grind a ...

2026-08-01T15:52:48Z

ran out of max credits on both Sol and Fable trying to grind a mathematical theorem and get some novel results

conclusions so far (thread):

if you work at OpenAI and want to help me make progress on ...

2026-07-30T02:19:48Z

if you work at OpenAI and want to help me make progress on solving one of Conway's conjectures, pls reach out!

i’m going to run out of my weekly Pro limit tomorrow but the progress so far has been encouraging, so i’d rather keep going

i’ve seen the academics program but i’m an amateur 🤷‍♂️

cool to see parallels to engineering process ...

2026-07-30T02:08:02Z

cool to see parallels to engineering process

https://eurosky.social/xrpc/com.atproto.sync.getBlob?did=did:plc:fpruhuo22xkm5o7ttr2ktxdo&cid=bafkreibt2qegjdybwh2vs7rkswmgmgoewkzlvsecsq4ku4v2mbpz7titym

how the internal team is doing #nevent1q…uv2g ...

2026-07-30T00:59:25Z

how the internal team is doing

quoting
nevent1q…uv2g

we’re trying to crack Conway’s conjecture that Oz is pre-Schreier

https://eurosky.social/xrpc/com.atproto.sync.getBlob?did=did:plc:fpruhuo22xkm5o7ttr2ktxdo&cid=bafkreidnlhxgko2djtn6qpknjz7ijyef7krmirg4ujfshry5v6x7wpnf24

https://eurosky.social/xrpc/com.atproto.sync.getBlob?did=did:plc:fpruhuo22xkm5o7ttr2ktxdo&cid=bafkreiedclgzbf3vpllqyddjk6kszmsaro67dvtj7rhdtsz4hva243umku

also i think it's amazing you can just ask them to explain ...

2026-07-30T00:45:47Z

also i think it's amazing you can just ask them to explain current step at any level of detail (i've suggested formalizing the critical path so that's a side quest of one of them)

https://eurosky.social/xrpc/com.atproto.sync.getBlob?did=did:plc:fpruhuo22xkm5o7ttr2ktxdo&cid=bafkreicjmk2bg5mp6tbvrhlxpij4qig7gqf3amkne2re53dsgup6xw2ldy

here's how my lil guys explain it part of my difficulty is ...

2026-07-30T00:33:01Z

here's how my lil guys explain it

part of my difficulty is that i struggle with the difference between prime and irreducible in general, which is a foundational topic. i need more of this stuff to settle in my head

https://eurosky.social/xrpc/com.atproto.sync.getBlob?did=did:plc:fpruhuo22xkm5o7ttr2ktxdo&cid=bafkreievuczm7k3lrgy2sd5zhjam3v5xg5mon2d4tdngaksp47fyjev6qe
https://eurosky.social/xrpc/com.atproto.sync.getBlob?did=did:plc:fpruhuo22xkm5o7ttr2ktxdo&cid=bafkreifyhui6bj5rvrinnnnbnvedw3lsp22badtmlrsuyviwqmycjt7hna
https://eurosky.social/xrpc/com.atproto.sync.getBlob?did=did:plc:fpruhuo22xkm5o7ttr2ktxdo&cid=bafkreighaa4qdgvmcekptcexp2g6rizbfojqiboiujj3h5ygtsxya46rsy

what we're trying to prove (or disprove) is this, if i ...

2026-07-30T00:29:52Z

what we're trying to prove (or disprove) is this, if i understand correctly. i don't know if this remaining half of the conjecture has a specific name

https://eurosky.social/xrpc/com.atproto.sync.getBlob?did=did:plc:fpruhuo22xkm5o7ttr2ktxdo&cid=bafkreifu7w4qiueryjhqdg7kid577gw45atl5z7hkckw6zbm3xsw5fnehi

another one about them www.infinitelymore.xyz/p/omnific-in... The ...

2026-07-30T00:27:48Z

another one about them www.infinitelymore.xyz/p/omnific-in...

The omnific integers are stran...

anyway, to your question — the conjecture is about omnific ...

2026-07-30T00:24:50Z

anyway, to your question — the conjecture is about omnific integers. they’re a subset of surreal numbers which are amazing. very good series on them here. this is the part about Oz www.infinitelymore.xyz/p/omnific-in...

The omnific integers are an in...

basically i don’t think it’s either/or. i want to understand ...

2026-07-29T23:55:16Z

basically i don’t think it’s either/or. i want to understand it but i also just want to flip it

but at the same time... i just like being in the middle of a ...

2026-07-29T23:45:13Z

but at the same time... i just like being in the middle of a storm of incomprehensible high level stuff i will probably never get the time to learn (or that is beyond my mental capacity) and it tickles me to use engineering management practices towards mining truths from the platonic realm?

i don't think you can say i don't like understanding. ...

2026-07-29T23:43:47Z

i don't think you can say i don't like understanding. i've read through the first 7-8 chapters of Tao's Analysis, and i did every. single. exercise. i've also done github.com/gaearon/anal... entirely by hand (although fewer of them). i like to "actually learn" things

GitHub - gaearon/analysis-solu...

the feeling to me is similar to being immersed in a language i ...

2026-07-29T23:42:14Z

the feeling to me is similar to being immersed in a language i don't understand yet. i love it. i'll maybe start recognizing some sounds and shapes but i just love being surrounded by it

i also really enjoy the "manager" role in the sense of helping these things actually do something useful

i know this isn't what people call metaprogramming but i ...

2026-07-29T23:36:59Z

i know this isn't what people call metaprogramming but i don't care. it's metaprogramming

metaprogramming is the new programming. half of what i do is like ...

2026-07-29T23:36:45Z

metaprogramming is the new programming. half of what i do is like "mutation test this", "confirm full coverage", "build a simpler model of this change and fuzz the hell out of it" etc

it's so cute they're roadmapping #nevent1q…uv2g ...

2026-07-29T18:58:05Z

it's so cute they're roadmapping

quoting
nevent1q…uv2g

we’re trying to crack Conway’s conjecture that Oz is pre-Schreier

https://eurosky.social/xrpc/com.atproto.sync.getBlob?did=did:plc:fpruhuo22xkm5o7ttr2ktxdo&cid=bafkreidnlhxgko2djtn6qpknjz7ijyef7krmirg4ujfshry5v6x7wpnf24

https://eurosky.social/xrpc/com.atproto.sync.getBlob?did=did:plc:fpruhuo22xkm5o7ttr2ktxdo&cid=bafkreia6vodehjnhtozkdgx3q3upngwhdobwpmn5brnsvivfk2n5ctdzle

i guess part of the motivation is just that there's a niche ...

2026-07-29T17:01:35Z

i guess part of the motivation is just that there's a niche in mathematics where i want progress to be made, but it is a niche so it is moving slowly. so i wonder if we can bruteforce some results there

part of it is also that (1) i like math and (2) i like Lean, but ...

2026-07-29T16:57:11Z

part of it is also that (1) i like math and (2) i like Lean, but (3) i know way too little. i think watching AI do this is similar to being an eng manager for a team making a product you don't understand that writes in a language you don't know

which is like.. silly but fun in its own way

i mean it wouldn’t be super interesting to me if there was no ...

2026-07-29T16:38:51Z

i mean it wouldn’t be super interesting to me if there was no way to certify whether result is correct. so this is why i’m doing a parallel track with a Lean formalization. which still doesn’t save you (need to have right definitions) but much easier to audit and constrain

added a separate Breakthrough guy whose only purpose is to come ...

2026-07-29T16:31:54Z

added a separate Breakthrough guy whose only purpose is to come up with breakthroughs

we’re trying to crack Conway’s conjecture that Oz is ...

2026-07-29T15:34:38Z

we’re trying to crack Conway’s conjecture that Oz is pre-Schreier

https://eurosky.social/xrpc/com.atproto.sync.getBlob?did=did:plc:fpruhuo22xkm5o7ttr2ktxdo&cid=bafkreidnlhxgko2djtn6qpknjz7ijyef7krmirg4ujfshry5v6x7wpnf24

Nostr event nevent1qqsqpzgwstm7cq5mtps7kewswddtgpmc8fex2e8cms25gumpdswnd0czyrat22ls2dd5tp0gy4syvwzxxteakz35h3we7jdg7qmclc47pp2zjm4d5pm

2026-07-29T13:18:27Z

(i just asked agents to do this)

made them a group channel ...

2026-07-29T12:06:02Z

made them a group channel

https://eurosky.social/xrpc/com.atproto.sync.getBlob?did=did:plc:fpruhuo22xkm5o7ttr2ktxdo&cid=bafkreievzhkxd5spy3zybzzcp63pcajdqe4olwk2inyr5rlfqoib55llwe

Nostr event nevent1qqsr02ed5gcsfm5anehcalslxvhjr4y7hqmv7f6cp7h7gcg5v76f82szyrat22ls2dd5tp0gy4syvwzxxteakz35h3we7jdg7qmclc47pp2zjjwedkl

2026-07-29T11:58:06Z

turning my codex into slack

my current setup is three Sol sessions - lab 1 attacks next ...

2026-07-29T08:52:07Z

my current setup is three Sol sessions

- lab 1 attacks next frontier of the problem
- lab 2 tries to formalize highest leverage piece in Lean
- the integrator pushes to main and keeps labs busy

i'll also add another lab for shower thoughts

i'm excited ...

2026-07-29T08:39:20Z

i'm excited

https://eurosky.social/xrpc/com.atproto.sync.getBlob?did=did:plc:fpruhuo22xkm5o7ttr2ktxdo&cid=bafkreic67etzciddcesoxrvzjtkmobqynxjwew3uy65e5kzaa6ejhzbtne

Sol has a very can-do attitude while Opus is grumpy but great at ...

2026-07-27T18:11:56Z

Sol has a very can-do attitude while Opus is grumpy but great at refining and poking holes

man Sol is so good. i’m doing some mathslop explorations, and ...

2026-07-27T18:07:02Z

man Sol is so good. i’m doing some mathslop explorations, and pasting replies between Sol and Opus 5 has been the best workflow so far, they can really cook together

this is most likely bs of course but i’m really enjoying ...

2026-07-27T11:57:37Z

this is most likely bs of course but i’m really enjoying Sol’s attitude over claudes. maybe first time i prefer openai model. it’s just.. unfazed

quoting
nevent1q…r3ay
lol this may actually check out. gonna have to spend more time on this

lol this may actually check out. gonna have to spend more time on ...

2026-07-27T11:49:14Z

lol this may actually check out. gonna have to spend more time on this

Nostr event nevent1qqs9ztasepk4cy46h9xpgul5zgr9fqyg3yq2qr7ewlmqw6mkyundtvszyrat22ls2dd5tp0gy4syvwzxxteakz35h3we7jdg7qmclc47pp2zjktnfq9

2026-07-27T07:24:21Z

which is why models good at one are good at the other

i think it works? ...

2026-07-27T03:05:16Z

i think it works?

https://eurosky.social/xrpc/com.atproto.sync.getBlob?did=did:plc:fpruhuo22xkm5o7ttr2ktxdo&cid=bafkreibukqbnkmp3w2lnh4sgiowlslrpkfsfe6kdwkz7hsywzvxbvnerau

proofs and security vulnerabilities are more similar than i ...

2026-07-27T02:31:46Z

proofs and security vulnerabilities are more similar than i realized. like you have to combine gadgets to escalate it

ah yes, the enemy’s infinite-dimensional freedom ...

2026-07-26T17:55:33Z

ah yes, the enemy’s infinite-dimensional freedom

https://eurosky.social/xrpc/com.atproto.sync.getBlob?did=did:plc:fpruhuo22xkm5o7ttr2ktxdo&cid=bafkreihlnfsjzkszhf62ojxlt4ntprk5je3epwlu67mpfjfgovims3lbdu

Nostr event nevent1qqsqjaqqy2q3ygwr4xm9uugrdx0yw2js5mp00z3xrhldfnvyv6vx2wszyrat22ls2dd5tp0gy4syvwzxxteakz35h3we7jdg7qmclc47pp2zjpk4rdj

2026-07-26T02:31:22Z

you can now distill codebases

declarative control over prefetching github.com/vercel/next.... ...

2026-07-23T14:15:18Z

declarative control over prefetching github.com/vercel/next....

https://eurosky.social/xrpc/com.atproto.sync.getBlob?did=did:plc:fpruhuo22xkm5o7ttr2ktxdo&cid=bafkreiep62koa4km4txxspkgcimdb4mgslnze673nnpit4lj6flete24hy

the point is just having good levers to do what makes sense for ...

2026-07-22T21:42:33Z

the point is just having good levers to do what makes sense for the product and the specific screen

whaaat #nevent1q…suam

2026-07-22T02:58:32Z

whaaat

quoting
nevent1q…suam
Just got a peek at bsky communities 👀

transcript from the end of the blog post ...

2026-07-22T02:07:16Z

transcript from the end of the blog post chatgpt.com/share/6a5fdc...

Check out this chat

some evidence that AI can be helpful in explaining AI findings ...

2026-07-22T02:06:18Z

some evidence that AI can be helpful in explaining AI findings terrytao.wordpress.com/2026/07/21/a...

A digestion of the Jacobian co...

quoting
nevent1q…ndfl
i’m optimistic about this. ai is pretty good at explaining things, removing unnecessary complication, refactoring etc. and i’m optimistic that interesting corners of math fundamentally “make sense” when broken down into explainable abstractions

literally me in the gym today #nevent1q…rdj9

2026-07-20T23:37:38Z

literally me in the gym today

quoting
nevent1q…rdj9

nevent1q…g6wk

https://elfcup.us-east.host.bsky.network/xrpc/com.atproto.sync.getBlob?did=did:plc:qwf4izoyeyesijdeyjlity7g&cid=bafkreigfwudi7ff66b4oym73cfozj5rj7dajpiy64iwivw222ctszyynvq

i’m optimistic about this. ai is pretty good at explaining ...

2026-07-20T14:41:13Z

i’m optimistic about this. ai is pretty good at explaining things, removing unnecessary complication, refactoring etc. and i’m optimistic that interesting corners of math fundamentally “make sense” when broken down into explainable abstractions

so my argument is — if we appreciate having those things ...

2026-07-20T14:33:21Z

so my argument is — if we appreciate having those things already mapped out, why is it a problem for a thousand years more of work to already be mapped out?

isn’t it what would have happened to mathematics a thousand years later anyway?

why would you NOT want to wake up in the future, with a thousands ...

2026-07-20T14:00:10Z

why would you NOT want to wake up in the future, with a thousands years’ worth of results, new theories and beautiful symmetries found, new objects and their connections? even if it’s many lifetimes to understand, why would you not want a glimpse at how deep the rabbit hole goes?

Nostr event nevent1qqswrgmkcp6gf9lyvvrwk4z0dhry50je26cf2ywcqzthv0jfsye94kczyrat22ls2dd5tp0gy4syvwzxxteakz35h3we7jdg7qmclc47pp2zj88nd70

2026-07-19T23:46:21Z

the urge to fix everything incorrectly