Do proactive agents really need an LLM to decide when to wake? The default proactive agent calls an LLM on every event just to decide whether to wake up. That is a lot of expensive inference spent on a yes or no. New research from Microsoft and Purdue asks whether the trigger really needs a language model at all. Their answer is a 220MiB temporal-graph encoder that decides when to wake and what context to anchor. It gains +16.7 mean F1 across 14 backbones, runs 4 to 83x faster, and fits on-device at around 11ms per event. If you run an always-on agent loop, the polling decision is quietly the main cost driver. A tiny encoder removes it without giving up accuracy. Paper: https://t.co/15KpQEm7Eo Learn to build effective AI agents in our academy: https://t.co/LRnpZN7L4c

// State-Externalizing Harnesses // A new paradigm is emerging on how to effectively build agents and harnesses. If there is a state that the environment can maintain reliably, it probably doesn't belong inside the policy. Move it into the harness, and a 20B model trains better and generalizes further. Search agents are usually trained on one policy over a growing transcript, so RL has to learn semantic search and routine bookkeeping at the same time. This model, Harness-1, splits those apart. The harness keeps the working memory (candidate pool, evidence links, verification records, deduplicated observations, budget-aware context) outside the policy, and the 20B model only decides what to search, what to keep, what to verify, and when to stop. Across eight retrieval benchmarks spanning web, finance, patents, and multi-hop QA, it reaches 0.730 average curated recall, beating the next-best open search agent by 11.4 points and staying competitive with much larger frontier searchers. The gains are largest on the held-out transfer. Paper: https://t.co/8DOQtsLsp2 Learn to build effective AI agents in our academy: https://t.co/LRnpZN7L4c

Nice primer on post-training reasoning data. (bookmark it) This is one of the first primers to pull the scattered post-training reasoning-data literature into one place, synthesizing over 150 public studies and system reports that previously lived across dataset papers, RL recipes, reward-model studies, benchmarks, and frontier reports. It organizes everything around four questions. What data objects exist, what makes them useful, how they are constructed, and how they scale. Paper: https://t.co/royylAHk3y Learn to build effective AI agents in our academy: https://t.co/LRnpZN7L4c

Outstanding paper on long-horizon agents. (bookmark it) Similar to humans, how do you make agents persist on a difficult task, and how is that useful? And which models today work well on this? This new work, AutoLab, explores this question and how encoding persistence in agents is beneficial for tasks such as auto research and engineering tasks. Can a model keep improving an artifact for hours, under a strict wall-clock budget, the way real research and engineering actually work? Results: AutoLab hands agents 36 expert-curated tasks across system optimization, model development, CUDA kernels, and puzzles, each starting from a correct but deliberately suboptimal baseline. Across 17 frontier models, the dominant predictor of success was not the quality of the first attempt. It was persistence, repeatedly benchmarking, editing, and folding in empirical feedback. It appears that Claude-opus-4.6 sustained that loop well. Most of the other models quit early or burned the budget, making almost no progress. Paper: https://t.co/jb8uYR0fpE Learn to build effective AI agents in our academy: https://t.co/LRnpZN7L4c

// Agents' Last Exam // Agents' Last Exam is a living benchmark of over 1,000 economically valuable tasks, built with 250+ industry experts and mapped to the U.S. federal occupational taxonomy. The hardest tier sits at a 2.6% average full pass rate across mainstream harnesses and backbones. ALE behaves like a GDP-coverage instrument instead of another test that saturates in a month. Paper: https://t.co/2FMeltJ23e Learn to build effective AI agents in our academy: https://t.co/LRnpZN7L4c

New research from Renmin University. Treat skill selection as a harness in its own right. If you design skill routing for personal or edge agents, this work argues that the selection layer is a first-class component you train and own, sitting alongside memory rather than inside it. The work builds a lightweight local preference harness for on-device personal agents. It keeps a cheap statistical preference learner on-device while a remote LLM handles semantic intent, and the local statistics modulate the model's skill-selection decisions rather than overriding them. Framed as a bandit-style local optimization, the decoupled design reports the lowest cumulative regret and highest test accuracy against memory-augmented agents. Paper: https://t.co/nBigS6jRf7 Learn to build effective AI agents in our academy: https://t.co/LRnpZN7L4c

Probably the most important AI-in-math paper to date https://t.co/P8gxSg6N0k

BPC-157 regrew a completely SEVERED sciatic nerve in 60 days. (PMID: 19903499) Does your back hurt before you get out of bed? Do you wince tying your shoes? Does pain shoot down your leg into your foot? That’s not “just back pain.” It’s nerve compression. This can lead to: → permanent nerve damage and foot drop → disc surgery with 40% failure rate → inflammation crushing your spinal nerves → muscle atrophy in your legs → losing the ability to walk pain-free BPC-157 also IMPROVED spinal cord crush recovery over 360 days (PMID: 31266512). A peptide your body already makes. Repairing what your back surgeon couldn’t. Advil shreds your gut. Cortisone breaks down collagen. Surgery fails 40%. This doesn’t. I take Barrier Health’s BPC-157 oral tablets personally. No injection. No prescription. Code ALFRED saves you 15%. Link below.

🏟️❤️ THE PARTY STARTED IN THE STANDS… AND ENDED ON TOP OF EUROPE! 🏆 19,000 Olympiacos fans turned the arena red and witnessed history as Olympiacos B.C. defeated Real Madrid Baloncesto 92-85 to become European champions once again! ⭐️⭐️⭐️⭐️ Scenes nobody in Piraeus will ever forget. Chants, tears, flags and pure emotion as the Reds celebrated their fourth EuroLeague crown. 🔥❤️🤍 THRYLOS FOREVER! 🇬🇷🏀 #Olympiacos #EuroLeague #Champions #FinalFour #Thrylos #Basketball #Piraeus #EuropeanChampions #GreekBasketball #Hellas

I’m happy to share INDUCTION: Finite-Structure Concept Synthesis in First-Order Logic, accepted to ICML 2026 as a spotlight. The benchmark aims to test models in their ability to generalize from examples: find compact logical rules that explain many given examples. Each problem gives several small finite relational worlds, with objects labeled as belonging or not belonging to a hidden concept. The model must output one first-order formula phi(x) that captures the concept. An important twist is that getting the examples right is not enough: models can sometimes fit the input worlds with large case-splitting formulas. We therefore score both correctness and formula size, and test whether formulas generalize to held-out worlds. The benchmark is designed to be feasible but not easy for frontier models. This table shows performance in three tasks. In FullObs, all facts are observed, and the formula must match the target concept across several worlds. In CI, the model sees YES and NO worlds. The goal is not to invert labels on the NO worlds, but to avoid formulas that exactly explain those contrastive worlds. In EC, some facts are unknown. A formula is valid if, for each world, there exists some completion of the unknown facts under which the formula matches the target labels. One main result is that validity can be misleading. Some model outputs are compact and close to the intended rule. Others are correct on the input worlds but use huge formulas with large disjunctions that branch on accidental properties of the finite structures. Those bloated formulas usually do not generalize. When we sample new held-out worlds from the same generator and label them by the planted rule, compact near-gold formulas generalize much better than bloated ones.

The arXiv preprint: https://t.co/wOYiFfSs4J The public repo: https://t.co/paZfTvZGZW

@TheCryptoCPA https://t.co/b774dXzDJ7

Claude Opus 4.8 is now available for Max subscribers on Perplexity and Computer. https://t.co/DNDNo0Iqxj

Today we're announcing that hybrid agentic inference is coming to Perplexity Computer. Computer can split tasks between a local model running on your machine and frontier models in the cloud. This keeps private data on your device and maximizes token efficiency. Coming soon. https://t.co/6t3PrmI1FX

Introducing "Dev" on Hermes Atlas - a tutorial for aspiring builders to understand how Hermes Agent works at a technical level This is the first step towards adding resources that will help you build with Hermes Agent Link in replies 👇 https://t.co/nzoSoG9dLE

🚨🌍World models are surprisingly fragile! We introduce BadWorld, an adversarial attack for visual world models. A tiny perturbation to the starting image 🖼️ can break down the whole world. Code:https://t.co/zsmljCDQoS Paper:https://t.co/XronO2Iq87 Arxiv:https://t.co/ELPwe3Gp3O https://t.co/VS4inrqNjv

+1 more

Made in London with AWS: Hirsh Pithadia, CEO & Co-founder, @ValyuOfficial. Valyu is building a search and retrieval solution for AI and it’s chosen London to do so. Pithadia discusses how the city’s close-knit startup community, research and engineering talent, and support from AWS Activate has helped PolyAI’s team launch and grow the company.

The welfare state has been more destructive to the black family than slavery just by restructuring the incentives. In 1960, nearly a century after emancipation, only about 22% of black children grew up in single-parent households. By 1990, after the Great Society welfare expansions, that number had more than tripled. Thomas Sowell has long shown this wasn’t the lingering shadow of slavery or some vague “legacy.” The destruction of the black family was the direct result of welfare policies that subsidized single motherhood and penalized marriage. The incentives changed, and family structure collapsed accordingly.

Meet Dreamina Seedance 2.0 Mini: Built for quality and speed, at lower cost. - ~30% lower cost than Dreamina Seedance 2.0 - 2× faster than Dreamina Seedance 2.0 Fast - Comparable output to Dreamina Seedance 2.0 Fast Launch offers: - Pro users (existing & new subscribed before 6.21) enjoy 33% fewer credit usage with Mini-720P from June 15 - July 22 - Up to 60% off Pro plan purchased via CapCut App This means Dreamina Seedance 2.0 Mini costs up to 55% less than Dreamina Seedance 2.0 with the limited-time offer. Generate more, spend less, keep the quality. Now rolling out across all relevant CapCut features: - App: AI Lab, AI Generator, AI Video - Web: Video Studio, Design Studio (now enjoy motion design with video models) - Desktop: AI Video, Edit Pilot RT+comment in 5hr to get 200 free credits in your DM.

Introducing Omma Studio, a full suite of AI tools to generate 3d, video/image, and audio with SOTA models. Everything made in Studio can integrate directly with your Omma's chat code projects to deploy/publish to the web. Content can be generated in parallel alongside your code generation in Omma, too. Studio -> Code -> Deploy

BREAKING: Starlink is delivering faster download speeds than other terrestrial ISPs in 22 out of 23 markets in Africa, per Ookla. • Upload speeds doubled in Central & East Africa and nearly tripled in Southern Africa. • Now serving ~500,000 users across rapidly expanding markets. • Latency reduced by over 80% in Southern & Eastern Africa since 2024. • 5 to 6 times faster than local providers in Botswana and Guinea Bissau Starlink is helping connect schools, businesses, homes and remote communities that were ignored for decades. It is becoming Africa’s internet lifeline. Meanwhile, South Africa continues to block it because of its racist BEE laws, denying its own citizens access to fast reliable internet.

SpaceX: $19 Mrd Umsatz in 2025, $5 MRD Verlust. Und das soll nun $2.519.000.000.000 wert sein? (2 Billionen 519 Milliarden)? Es ist die größte Finanzblase in der Geschichte der Menschheit. Dümmer als Tulpen. https://t.co/kDpyxxOtvl

Today, we're announcing Factory 2.0: from coding agents to software factories. https://t.co/yBbirv4Ik9

Introducing DreamX-World 1.0 — a general-purpose world model with 1 minute continuous generation, real-time interaction, precise camera control & multi-style support. Beta coming soon! 🌐https://t.co/U5NIvOi6rU Github：https://t.co/zolKYnuWjT #WorldModel #AIVideo https://t.co/3xbGJn0Va9

@adilmania New media: https://t.co/kiuZ7QXLzb :-)

everyone's upset that fable 5 is gone, but can i remind you that opus 4.8 just one-shotted this: an AI badminton scorekeeper point a phone at the court → it follows the shuttlecock, keeps score, and tells you how hard you're whacking it get back to building friends https://t.co/Kj6A9PafLP

This AI just exposed the BIGGEST legal insider trading operation in America. A platform called GovGreed built a seven-layer machine learning system that cross-references every stock trade disclosed by every sitting politician against the bills their committees control, the campaign donations they receive, and the companies their votes directly impact. It scored all 540 politicians currently in Congress. And the numbers are crazy: 56% of every stock purchase made by Congress in the last 16 months was on a stock directly affected by a bill the buyer later voted on. That is 6,170 out of 11,016 total purchases. More than HALF of all congressional stock buys are on companies whose fate that same politician is about to decide. 343 of 540 Congress members actively trade stocks while holding access to nonpublic legislative information. That is 63.8% of the entire legislature making market bets with an informational edge that would put any hedge fund manager in prison. The AI identified 752 active "Triple Signals" in the current Congress. A Triple Signal fires when three conditions line up at once: The politician sits on the committee controlling a bill, they traded stock in a company affected by that bill, AND they received campaign contributions from that same industry. Bills carrying these insider indicators pass at 5.4 TIMES the normal rate. Now look at the individual leaderboard: - Nancy Pelosi's estimated portfolio sits at $194 million with a Greediness score of 98.1 out of 100 - Ro Khanna made 13,231 trades across 800+ different tickers - Michael McCaul made 32,302 trades and filed 6,670 of them late - Thomas Suozzi filed 86.4% of his trades late with an average delay of 396 days, meaning his disclosures landed over a YEAR after he made the trade And then there is Lisa McClain, the fourth-ranking Republican in the House. She has made 1,443 trades in three years, more than 98% of all politicians tracked. She violated the STOCK Act twice in a single year, disclosing up to $900,000 in trades months after the legal deadline. Her husband bought up to $250,000 in Elon Musk's xAI, which quietly converted into SpaceX equity before last Friday's $2 trillion IPO. The penalty for all of this? A $200 fine. The number of Congress members ever prosecuted under the STOCK Act since it passed in 2012? Zero. And the cruelest part is this: A bill to ban congressional stock trading was introduced in January 2026. It has bipartisan support. Over 80% of American voters want it passed. But Congress is sitting on it, because the people who would have to vote yes are the same people making millions from the system staying exactly the way it is. They write the insider trading laws, they exempt themselves from enforcement, they trade on the information those laws generate, and when they get caught, they pay a fine that is basically nothing. The AI didn't discover anything Congress was hiding. It just organized what was already public into a pattern so obvious that nobody can pretend it isn't there anymore.

not since tulipmania in 1636 have things been this loopy. https://t.co/CZ83Ge9R0k

The Trump administration is looking to take financial stakes in OpenAI & Anthropic. The recent Anthropic Claude Fable 5 & Mythos 5 restrictions could be a leveraging tool: "Maybe this could be kind of a negotiating tactic…” — @steph_palazzolo, AI Reporter https://t.co/3ZIgGklZ8m

not since tulipmania in 1636 were things this loopy. https://t.co/WRnuEFE5Xg