Great talk exploring open-ended ness. My favorite from the summit.
ARC Prize Research Summit - @kenneth0stanley
On why open-endedness, invention, discovery are essential to AGI, but designing a benchmark for them creates difficult problems.
0
0
3
247
Absolutely brilliant idea and execution from @guille_bar. I pitched in on the packaging design. So fun! 🎈✨
I got a fun birthday present from the team at Ndea ❤️
0
0
5
210
Got to MC this one. Great speakers. Fun (and smart) room.
ARC Prize Research Summit 2026
Last Friday we held our first ARC Prize Research Summit @ MIT in Boston
We enjoyed presentations from @stephen_wolfram, @kenneth0stanley, @sirbayes, @alexisfox and Alexandre Bouayad
Thank you to everyone who attended
Recordings coming soon
1
0
8
451
Cool to see this ship. A friend has been working on a foundation model for human behavior. Makes agent training + evaluation more realistic. Go, Kevin!
Today we introduce Kairos 1: the world's first model built to simulate individual human behavior. Frontier models only understand population averages, ignoring the individual differences that make us human.
Kairos beats Persimmon, Astra, Fable, and Gemini across 13 simulation benchmarks.
1
0
5
317
Bryan Landers retweeted
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
Application deadline: September 14th, 2026, 11:59pm anywhere in the world
Travel expenses will be covered for selected presenters
ARC Prize Research Summit 2026
October 23, 2026 - Boston, MA
The ARC Prize community comes together to share new work, challenge ideas, and advance the frontier of machine reasoning towards general intelligence
3
6
31
6,724
Bryan Landers retweeted
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:
- 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:
133
408
3,613
1,184,229
This fall in Boston - top researchers from MIT and around the world will meet to explore the state of ARC.
We'll talk program synthesis, neurosymbolic AI, world models, agent harnesses, reasoning benchmarks...and how we get from here to AGI.
Join us!
We'll talk program synthesis, neurosymbolic AI, world models, agent harnesses, reasoning benchmarks...and how we get from here to AGI.
Join us!
ARC Prize Research Summit 2026
October 23, 2026 - Boston, MA
The ARC Prize community comes together to share new work, challenge ideas, and advance the frontier of machine reasoning towards general intelligence
0
1
10
1,052
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.
3
4
47
6,117
Turned the tables on MLST's @ecsquendor to interview him on the Ndea pod along with his paper co-author, Jeremy Budd.
Tim's content has inspired the ARC world (and beyond) for years. His quality and use of AI agents in media production is unmatched.
Definitely watch this one!
Tim's content has inspired the ARC world (and beyond) for years. His quality and use of AI agents in media production is unmatched.
Definitely watch this one!
On the pod: @MLStreetTalk's paper "Why Creativity Cannot Be Interpolated" with @ecsquendor + Dr. Jeremy Budd.
A deep dive into creativity, understanding, neuro-symbolic AI, human-AI collaboration, and how systems capable of extending their own phylogeny could become creative.
1
1
5
420
🚀
Current models rely on brute-force RL training to adapt to new tasks and environments. However, intelligent systems should be able to learn on the fly in a sample-efficient manner, continuously refining their understanding of the world through their interactions.
After Labs is developing models with built-in adaptation mechanisms rather than treating adaptation as an afterthought. This new generation of models will start a new era of curious agents that may not possess all the world's knowledge but will have the crucial ability to learn in real time.
Today, we are happy to announce that After Labs has been selected as one of 10 AI labs to join NFAI and receive €3M in additional funding. This support will accelerate our mission to build models that continuously learn about the world. We will be contributing to open science along the way, so stay tuned and follow us on our journey to develop AI that augments human understanding of the world 🌍
1
0
3
340
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
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
Today, we’re introducing [schema]: a harness reaching 99% RHAE with Opus 4.8 + Fable 5 and 95.35% with GPT-5.6 Sol on ARC-AGI-3 Public set.
[schema] makes an LLM think like a physicist. 🧵
7
17
224
21,922
I finally infiltrated the elite program synthesis department at UCSD. 😎
Been a pleasure getting to know @lorisdanto and find ways to collab. Excited to learn more about the PL group’s work.
Been a pleasure getting to know @lorisdanto and find ways to collab. Excited to learn more about the PL group’s work.
On the pod: "Constrained Adaptive Rejection Sampling" with @ucsd_cse professor @lorisdanto.
Hear how symbolic AI experts have navigated the LLM era and why the future of AI code generation depends on program synthesis and formal methods.
2
1
11
1,356
Successful pre-AI startups often had minimal customer-facing surface area:
Stripe was an API.
Brex had no UI, mostly terminal/manual flow.
Airbnb was a simple form.
DoorDash - same.
The danger with AI is that it becomes too easy to build too much. Founders lose discipline about what actually matters. But intelligence is compression.
"If you can't minimize your surface area and and solve the problem with a very clear set of boundaries, you haven't found the right problem to solve."
@pedroh96 on YC pod.
Stripe was an API.
Brex had no UI, mostly terminal/manual flow.
Airbnb was a simple form.
DoorDash - same.
The danger with AI is that it becomes too easy to build too much. Founders lose discipline about what actually matters. But intelligence is compression.
"If you can't minimize your surface area and and solve the problem with a very clear set of boundaries, you haven't found the right problem to solve."
@pedroh96 on YC pod.
1
0
4
550
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
We’re going to need the equivalent for the agents that represent your brand
2
1
18
1,192
Tried OpenClaw - super slow even for dumb test requests like ‘Reply with “hi”’. Set up Hermes - fast. Easy decision so far. Having fun.
0
0
3
248
Bryan Landers retweeted
Good design is the art of packing 1,000 "hows" into a single "what". Good design is compression: making the numerator trend towards infinity while the denominator stays at 1.
29
18
333
31,423
Does intelligence require dreaming?
Absolute blast to have the great Kevin Ellis on the Ndea podcast!
Multiple people I've spoken to for the pod (and even one on our team) got into symbolic AI because of Kevin's "DreamCoder" paper. Legend.
Absolute blast to have the great Kevin Ellis on the Ndea podcast!
Multiple people I've spoken to for the pod (and even one on our team) got into symbolic AI because of Kevin's "DreamCoder" paper. Legend.
On the pod: our most-requested guest! @ellisk_kellis from @Cornell shares the origins of his influential neurosymbolic paper "DreamCoder".
Plus: program synthesis, wake-sleep library learning, world models, running an AI research lab, and more.
1
0
7
434