Co-Founder @BuildCoherence | Building Axiom to elevate logical discourse and knowledge 🚀 | Husband of @RealMetaBias
Joined February 2026
- Tweets 362
- Following 51
- Followers 41
- Likes 748
40 Photos and videos
Pinned Tweet
May 16
In Axiom, arguments are proven valid in Lean. Here, if you accept the 6 premise nodes, you must accept the conclusion.
The point is not that this is a perfect argument, it's that we can use this imperfect argument to pinpoint where our reasoning needs to be improved. Readers in Axiom flag which premises they disagree with and the author can then work on either changing their argument to be about a weaker claim (e.g. "We mustn't eat factory farmed meat"), or deriving their original premises as a conclusion from less controversial premises.
ALT A flowchart using 6 gray premise boxes and 2 purple conclusion boxes to argue against eating meat. Top row (solid gray): P1 (duty to perfect moral character), P2 (mistreating animals corrupts character), P3 (exploiting for commercial gain is mistreatment), P4 (eating meat exploits for gain). P1 and P2 point to a dashed gray 5th premise (P1∧P2→C1), leading to purple C1: "We mustn't mistreat animals." C1, P3, and P4 point to a dashed gray 6th premise (C1∧P3∧P4→C2), leading to the final purple C2: "We mustn't eat meat."
1
2
116
Programming languages will be re-released as a novel discovery in two years, but they will be called specification languages and will probably use dependent types.
Code is too tedious but prompting is too chaotic. The middle ground sounds reasonable to me, but it would be funny if this pseudocode representation of your software doesn’t just end up in an AI lab re-releasing programming languages as a novel discovery in two years.
1
AI agents being good at writing code doesn't mean programmers won't have jobs, it means everyone will have to be a programmer.
4
Sam Cymbaluk retweeted
Jun 10
Natural language has increasingly approximated formal logic over time.
The end state is that the base language is formal logic.
1
2
18
Jun 10
The fish doesn't know about the water it swims in; you probably don't know about the flaws in natural language that are holding you back.
8
Jun 10
Am I the only avid AI agent user that doesn't even think about benchmarks? 67% vs 71% on DeepSWEBench Diamond?! Does this mean something to someone?
Fortunately at this point, my brain mostly filters out benchmark charts, as they've never proven helpful in informing my actions.
48
Jun 5
It's interesting to think about what this means for social media platforms.
15
Sam Cymbaluk retweeted
Jun 5
I've got bidirectional feedback like this working with just md files. We have a review gate that forces compliance to our ontology. If an agent's PR contradicts the upstream docs, it is forced to update those docs. So a code change flows upwards, making review easy.
1
1
1
23
Jun 3
AI won't replace you, not because layoffs due to AI aren't real, but because replace is the language of a zero-sum substitution. It reflects a fundamental misunderstanding of the transformation we're undergoing.
9
Jun 2
Fitting branch name for work on formalizing language. I think Newton would adore how far humanity has come since he started us down this path.
1
9
Jun 1
There's a shocking difference in CO2 levels running a "portable" air conditioning unit in one room of our house. Modern AC installations focus purely on energy efficiency, but the tradeoff is real. For my family, fully functioning brains are worth the extra energy cost.
17
Sam Cymbaluk retweeted
Jun 1
Good thread of agent failures. How these ground facts translate to market impact is the big question. Agents autonomously writing complex code doesn't mean all coding jobs are done for. Conversely, bad AI failures don't imply there won't be lots of coding jobs made obsolete.
1
9
May 31
This is probably the most magical agentic coding setup I have working so far. Tag an issue and an hour later a senior quality task or bug fix PR is ready for review. It's grounded in our in-repo docs and often finds other areas for improvement beyond the issue scope.
3
35
May 31
I'm seeing a lot of negativity about AI agents for coding. Some of the hype pullback is healthy; delusional optimism is bad. But don't lose sight of the magic!
Yes, we are all now very familiar with all the ways agents fail, but the kind of things they CAN do are incredible. Go deeper, use the failures to understand the tech better, and reinforce the patterns that work.
5
5
236
May 31
I haven't noticed much of a difference between Opus 4.7 and 4.8 in claude code, but 4.8 in regular chat is dumb as a stump. I feel like I'm back in 2024. It's so bad I assume something must be broken with the CoT (I'm on high thinking mode).
1
3
65
May 28
Natural language puts all the burden on the reader to reconstruct the logical structure the writer had in mind. That effort is wasted when the writer didn't actually have anything coherent in mind. Historically, eloquent prose was a decent signal of worth because few people were both smart enough to write eloquently and foolish enough to spend their time writing and publishing an incoherent thought.
The solution is to make the logical structure itself the direct communication medium. All the hard pieces needed to do this effectively are there, someone just needs to put them together in the right way.
May 28
The reason it’s bad to post AI writing as your own isn’t quality, but economics. AI enables you to take a clumsy half-formed thought and turn it into something that sounds coherent and scholarly for pennies. The problem is that the effort required to vet your claims now exceeds the effort required to produce them. If you just posted the clumsy half-formed thought in your own words, it would be easy to detect and dismiss it as worthless.
I am strongly pro-AI. I welcome AI agents posting openly as AI agents. But messages from humans carry an implicit cost signal, and plagiarizing from AI undermines this.
Write your own content.
2
37
May 27
You're killing me @claudeai code web. In the past few days it switched to shallow clones and they simply don't work.
17
May 27
"Our enemies are not bad people – but rather bad ideas." ≡ "For we wrestle not against flesh and blood"
May 27
Dans le manifeste "techno-optimiste" de Marc Andreessen, il y a une phrase qui m'a marqué :
"Our enemies are not bad people – but rather bad ideas."
Nos ennemis ne sont pas des mauvaises personnes. Ce sont des mauvaises idées.
Prenons Jancovici. L'homme est brillant, sincère, travailleur. Il ne se lève pas le matin en se disant qu'il va nuire à l'humanité. Mais l'idée qu'il porte la décroissance, le rationnement, la frugalité érigée en horizon civilisationnel est une idée profondément destructrice. Elle prend des esprits brillants et les transforme en commissaires politiques d'un futur appauvri.
Et le plus fascinant, c'est ce que cette idée fait aux gens qui l'adoptent.
Dans mon entourage, une grosse partie de mes amis est sur cette ligne décroissantiste, avec tout le package qui va avec. L'argent c'est mal mais ils en veulent. Il faut moins prendre l'avion mais ils rêvent de voyager partout. Il faut consommer moins mais ils ne renoncent à rien de ce qu'ils aiment vraiment.
Et tous ont un point commun : ils sont déprimés. L'un d'eux m'a même confié qu'il était sous antidépresseurs.
Ce n'est pas un hasard. C'est mécanique.
Quand tu crois que ton désir de vivre, de créer, de t'élever est moralement suspect tu te détruis de l'intérieur. Tu passes ta vie à t'excuser d'exister. Tu vis dans la dissonance permanente entre ce que ton corps veut (plus, mieux, plus loin) et ce que ton idéologie t'ordonne (moins, sobre, immobile).
D'où ma théorie :
Quand on pense quelque chose de fondamentalement faux décroissance, communisme, extrémisme religieux (de tout ordre) ce n'est qu'une question de temps avant que ça devienne vraiment destructeur.
D'abord pour soi. Puis pour les autres.
Les mauvaises idées tuent. Lentement chez ceux qui y croient, brutalement chez ceux qui les subissent.
C'est pour ça que la bataille des idées n'est pas un luxe d'intellectuel. C'est la bataille la plus importante de notre époque.
10
May 26
This is great, but code is about the what, not the why. If we want to collaborate effectively with agents on a sufficiently complex why, we need formal logic as an agent harness.
May 25
This Meta Stanford Illinois survey paper argues that AI agents work better when code becomes their main working layer.
The problem is that an LLM by itself is mostly a text predictor, so long tasks can lose state, hide mistakes, and turn plans into actions in fragile ways.
The real advance is not “AI writes code,” but “AI uses code as the environment it thinks inside.”
The authors call the surrounding system an agent harness, meaning the tools, memory, sandboxes, checks, and feedback loops that turn a model into an agent.
Their core idea is that code should sit at the center of that harness, because code can be run, inspected, checked, saved, edited, and shared.
Tests become sensors.
Repositories become memory.
Logs become history.
Sandboxes become boundaries.
A generated script is no longer merely an answer; it is a handle the system can run, check, revise, share, and roll back.
The main finding is a pattern across many fields: code helps agents reason through executable steps, act through tool calls or control programs, and model environments through tests, traces, logs, repositories, and simulators.
----
Paper Link – arxiv. org/abs/2605.18747
Paper Title: "Code as Agent Harness"
1
40
May 25
"alternative, highly effective neutral strategies"
a.k.a. Logic
May 9
A new study shows that individuals with similar levels of autistic traits are naturally drawn to one another. Brain scans reveal they use alternative, highly effective neural strategies to connect, challenging traditional deficit-based models of autism. dlvr.it/TSSN7t
28