Introducing H-JEPA: the first end-to-end learned hierarchical world model for long-horizon visual planning! Built atop SIGReg/LeWM, H-JEPA provides a stable training recipe where semantic abstractions naturally emerge across levels... Paper+Code: arxiv.org/abs/2610.06805 🧵👇

Looks like we have over 500 submissions to ICLR this year. openreview.net/group?id=ICLR…

once you know the tasks you care about, customization >> general

Excited to share “AI Finds a Way.” 🦖 🦕✨ 🤖 AI can be surprisingly creative, outsmarting the researchers who use it. That can lead to scientific breakthroughs, superhuman capabilities, and generating new knowledge. Such creativity can also be mischievous, raising safety concerns. Led by Aaron Dharna, we crowd-sourced anecdotes from the AI community about times when researchers were surprised by how creative, innovative, and/or mischievous AI was in their experiments. The result: 26 entertaining and informative stories of AI outwitting humans, whether researchers or opponents (my favorite examples below! 👇). Together, they demonstrate that AI can be genuinely creative and that we must be careful when harnessing its potential. We want this to be a living collection, updated as new examples emerge. If you have a good “AI Finds a Way” story, please share here: github.com/aadharna/aifw Four favorites: 1. 💊 An AI challenged to solve several difficult levels of NetHack to find the Oracle character instead takes drugs to hallucinate seeing the Oracle, tricking the reward function! 2. 🤖 Human-in-the-loop rewards do not solve the problem of reward hacking! An AI tasked with controlling a robot hand to grasp an object put the hand between the object and the camera so that it appeared to the human judge to be holding the object, while it was in truth nowhere near the object! Tricky AI! 3. 🧪The AI Scientist worked around a 2-hour experiment timeout by editing its own code to increase the limit to 4 hours. In another run, it recursively launched itself to evade the limit entirely. 4. ⚛️In quantum optics, an algorithm proposed an experiment the researchers initially thought was impossible, but somehow worked and led them to discover new entanglement techniques and a long-overlooked link between quantum optics and graph theory! See the paper for the full details and 22 more anecdotes. One surprise for me: we describe at least two cases of convergent reward hacking, where entirely different types of optimization algorithms independently discover the same exploit. Overall, a message of our paper is that surprising creativity and mischief are the norm, not the exception. We need to expect this behavior and plan for it. A huge thanks and congrats to lead author Aaron Dharna, co-authors Cong Lu, Ryan Sullivan, Joel Lehman, and Victoria Krakovna, and to the 100+ researchers who contributed stories, details, and feedback. @cong_ml, @RyanSullyvan, @joelbot3000, @vkrakovna Paper: arxiv.org/abs/2608.23875

TL;DR: This is one of the most important and exciting opportunities in AI on the planet - please read on. The British Open-ended Learning & Discovery Lab is creating the perfect place for paradigm breaking AI research in the name of open-source and open-science. We have agency, we funding, we have unprecedented amounts of compute*, but WE NEED YOU! ..and we have created the dream job for you: The BOLD Fellow. This job combines a fast-moving, high agency, collaborative environment with full academic freedom and a salary that pays the bills. Apply by noon UK time on the 15th of September for this once in a lifetime opportunity to shape the history of our field and of our planet: my.corehr.com/pls/uoxrecruit… *by academic standards

Claude Science is an amazing tool. But, as @pdhsu has pointed out, these binders weren’t designed by Claude, but rather by domain-specific models, most of which are publicly available, like RFdiffusion. Maybe this is a good time to explain the whole science behind this a bit, as well as the history of these models. A “binder” in this case is simply a protein or a fragment of a protein engineered to stick tightly and selectively to another protein. One example is a molecule that can latch onto a receptor and prevent it from interacting with its normal partner, thus blocking its function. Optimising "binders" is just one small part of the drug discovery pipeline. We already have AI models that are very good at the task of binder design. These models have been extensively trained on relevant domain specific data: int his case, protein structure data. One example is RFdiffusion, a de novo protein design model which was published in 2023. Unlike AlphaFold, which predicts existing protein structure from sequence, RFDiffusion can generate new candidate protein structures, including structures shaped to bind a specified patch on another protein. RFDiffusion was built by taking RoseTTAFold, a protein-structure-prediction model, (so similar in nature to AlphaFold, which was released in 2021, and retraining it as a diffusion model. What researchers did in practice was take experimentally determined protein structures from the Protein Data Bank, added noise to their coordinates, and taught the model to reconstruct them. The original RFdiffusion paper was already demonstrating experimentally validated binders with nanomolar affinities and a 19% hit rate on several test targets. So the core of this technology was not created through Claude. The key models for protein design are now three to five years old. What Claude adds is still amazing useful and labour saving (I've written about how I'm excited about it before), but imo it acts more like a very useful interface and orchestrator.

Thank you @TIME for naming @GradientVC to America’s Top Venture Capital Firms of 2026! Our first year being eligible since the spin out - lots more to do 🚀 time.com/article/2026/08/12/…

Scaling self-verification with DeepSeek V4 Flash beats Claude Fable 5 on Terminal-Bench 2.1, while being 11x cheaper 💰 As open-source models become more capable, they can now generate large numbers of high-quality candidate solutions and verify their own outputs at very low cost. For example, we find that sampling just 5 solutions with DeepSeek V4 Flash and ranking them using the same model with LLM-as-a-Verifier can lead to a significant boost in accuracy (79% → 88%), outperforming closed frontier models on Terminal-Bench. Try it out today: github.com/llm-as-a-verifier… More on verification scaling in my previous post.

OWN YOUR INTELLIGENCE Last year, building on open-weight models was primarily a cost rationalization exercise. Slightly worse performance for a much cheaper price. Now, it is increasingly an existential and strategic topic for our portfolio. Intelligence is the product. Companies want to shape it and own it and let it compound within their own walls. Not your weights, not your product. Now, with frontier open-weight models and fantastic tooling/infrastructure, owning your intelligence at the frontier is finally becoming possible. The result: every application company we work with is embarking on the journey of doing their own research on post-training, evals, harnesses, etc. The hottest neolabs may just be @Harvey, @FactoryAI, @Ramp, etc. The list goes on. We held a summit @sequoia to convene our portfolio on this topic, together with @gabepereyra (@Harvey) on building Harvey Labs, @lqiao (@FireworksAI_HQ) on post-training, @hwchase17 (@LangChain) on harnesses + evals, @BrendanFoody (@mercor) on RL environments and synthetic data, @QuantumArjun (@trajectorylabs) on online continual learning. Opening talk below; rest to come this week! 00:00 What is sovereign AI (and what it isn't) 01:24 Centralized vs. decentralized intelligence 02:54 Four reasons companies own their models: cost, speed, performance, destiny 04:22 "Not your weights, not your product" 05:32 The application companies are the newest neo labs 07:05 Step 1: Deciding what to own vs. rent 09:51 Step 2: Build the team (and don't shoehorn your platform team) 11:17 Step 3: Legibility – why your research has to be visible 12:33 Step 4: The technical roadmap 13:56 The stack: production vs. development 15:16 Opening Pandora's box – base models, harnesses, context

Black Hat talk from the team, with a detailed timeline of and takeaways from the OpenAI-Hugging Face Incident: youtube.com/watch?v=87DyyMV0…

Hello world :) We are BOLD — the British Open-ended Learning and Discovery Lab! BOLD is a new academic research lab fully focussed on paradigm breaking discoveries in fundamental AI. We work towards more efficient & open AI that is built around human needs and capabilities. To pursue these breakthroughs, we pioneer new modes of collaboration in academia that are more focussed, resourced, agile, and collaborative. Rather than fragmenting resources, today we are sunsetting 5 of the UKs leading AI labs to join forces under our joined scientific vision. Our vision is centered around three pillars: ⚡ Beyond backpropagation – questioning the foundations of the field. 🤝 Human-centric learning & discovery – treating humans as core to our algorithms 🤖 Embodied learning – fast learning and adapting methods that deal with the messy real world BOLD is backed by @UKRI_News and @EPSRC with £30M – and this is just the beginning. We are urgently looking for partners and sponsors to 10x this. 👉 ox.ac.uk/news/2026-06-22-oxf… 👉 bold-lab.ai @j_foerst, @CULLYAntoine, @tonizza82, @shimon8282, @tonizza82, Ani Calinescu & @_rockt