Final call for applications to present at ARC Prize Research Summit in Boston on Oct 23rd
Application deadline: September 14th, 2026, 11:59pm anywhere in the world
Travel expenses will be covered for selected presenters
GPT-6 Astra by @OpenAI achieves SOTA on ARC-AGI:
- Astra scores 63% on ARC-AGI-3, 99% via a new provider adapter harness
- It surpasses human performance on 96% of ARC-AGI-3 levels
- It builds the most precise symbolic model of novel environments we've seen
Our analysis:
Bryan Landers retweeted
Excited to start filling in the left half of this chart, especially with open weight models
Inkling Small from @thinkymachines on ARC-AGI (Verified):
- ARC-AGI-2: 40.1%, $0.23/task
- ARC-AGI-1: 84%, $0.11/task
Inkling Small is the highest-scoring open-weight model evaluated by ARC Prize on both ARC-AGI-1 and ARC-AGI-2, setting a new cost-performance frontier.
Bryan Landers retweeted
Wow another one
The amount of research built on top of ARC-AGI-3 coming out is starting to snowball
The timeline is +3.5 months since V3 launch
Looks cool - need to dig into it
Bryan Landers retweeted
Your “company values” are an attempt at a system prompt for your employees
We’re going to need the equivalent for the agents that represent your brand
Bryan Landers retweeted
If you build an automation machine, the way to monetize it is to sell it to as many people as possible -- anyone who has tasks to automate.
But if what you build is an invention machine, then the best way to monetize it is to use it yourself.
Bryan Landers retweeted
Replying to @lulumeservey
i am but a single woman in tech, but my tech career started off by being forced to participate in a marketing video that made an oblique reference to semen on my (blue) dress, i watched men spy in interview rooms of female engineers to see if they were "hot" enough to hire, interns expected to participate in frat drinking culture then preyed upon by senior leaders, my own ass grabbed and being told "what? i like you" by a founder of the company, finding out that my own ($32k / year!!!!) was less than every other man on team w same role, told that i would be insane to have kids and be a founder and thank god i wasn't doing that (i was 6 weeks pregnant at the time), had an AE complain that I (the CPO of the company!!!!) was "too busy playing with my baby" to travel to *his* deals even though i returned from mat leave at 4 weeks and even took my 1 year old to vegas so i could keynote our user conference, watched every woman in my department but me get laid off and when i called it out was told "well but you know all of them were crazy," have been THE ONLY woman in almost every board room i've been in (both on operator side and board member side), get called "that broad" when i give a talk on some of the biggest stages in tech, may i go on.
i have done well for myself DESPITE all this.
i'm raising boys and don't think the vilification of men has been healthy in many aspects of our industry and culture, but I don't think my experience is unique so calling it "DEI fanfiction" is a BIT MUCH.
Bryan Landers retweeted
I had a great time discussing how problem decomposition can improve the performance of symbolic program synthesis systems. Hope you enjoy the episode! Big thanks to @bryanlanders for having me on!
"...simply looking at a problem differently can greatly improve learning performance."
New pod: Ndea researcher & ILP expert @CelineHocquette on "Relational Decomposition for Program Synthesis" - making symbolic learning systems more efficient without domain-specific knowledge.
Céline Hocquette on decomposition + ILP
Excerpt from an Abstract Synthesis podcast episode
Bryan Landers retweeted
Super exited that our recent work got featured in the Abstract Synthesis podcast! 🚀
I joined Brian to discuss inductive reasoning in the visual domain and how we can combine Vision-Language Models with Program Synthesis to enable more reliable and interpretable reasoning 💡
Vision-language models (VLMs) can see well, but they struggle to reason.
In this episode, @toniwuest (PhD researcher, @TUDarmstadt) explains how combining VLMs with program synthesis yields more reliable visual reasoning, with fewer tokens than chain-of-thought.
We need programs, not just prompts
A neuro-symbolic approach to visual concept induction that treats VLMs as perceptual tools inside symbolic programs.
Bryan Landers retweeted
Interesting interview! Great work by @bryanlanders on the interview. He did so well with the ARC interviews that I thought he should do a podcast!
Super excited to launch the @ndea podcast - Abstract Synthesis! I'll be interviewing leaders in the space of program synthesis and symbolic AI.
First up is the great @MarkSantolucito - the perfect guest to set up context for the new pod.
Lmk what you think!
Bryan Landers retweeted
The field of program synthesis is growing but is still under-producing the amount of talent I think we'll need to craft future AI systems. Our goal with Abstract Synthesis is to steer interest by highlighting frontier symbolic research and useful open problems.
Introducing the Ndea podcast - Abstract Synthesis. Hear the stories behind interesting academic papers in the world of program synthesis.
Episode 1 features @MarkSantolucito, @BarnardCollege/@Columbia, discussing his paper "Grammar Filtering for Syntax-Guided Synthesis".
Bryan Landers retweeted
ARC Prize 2025 Winners Interviews
Paper Award 3rd Place
@LiaoIsaac91893 shares the story behind CompressARC - an MDL-based, single puzzle-trained neural code golf system that achieves ~20–34% on ARC-AGI-1 and ~4% on ARC-AGI-2 without any pretraining or external data.
Bryan Landers retweeted
ARC Prize 2025 Winners Interview Series
Paper Award 1st Place
@jm_alexia shares details about the Tiny Recursive Model (TRM) - a ~7M-parameter, single-network recursive model w/ separate answer and latent states that attains ~45% on ARC-AGI-1 and ~8% on ARC-AGI-2
Bryan Landers retweeted
This year we did interviews with the Top 3 from both Paper Prize + Top Score
This is a ton of work! But I'm glad we have the history
Shoutout to @bryanlanders for getting all these done
Bryan Landers retweeted
The @arcprize discord is active, has great conversation, but needs love
spam, low quality posts (many of which look like chatgpt induced psychosis)
We need a part-time mod for it
Who's up for a paid community moderation role?
Bryan Landers retweeted
We're co-hosting a party with @ycombinator at NeurIPS this year
I'll be there along with @mikeknoop @fchollet @bryanlanders @agupta and Francois Chaubard
San Diego Rooftop with food/drinks
NeurIPS Party - ARC Prize Foundation + Y Combinator
Join us in San Diego for a NeurIPS party co-hosted with @ycombinator
Meet ARC Prize and YC leadership, researchers and industry leaders pushing the boundaries of frontier AI
San Diego
6-8 PM, December 6, 2025
























