AI News

OzBrain is a shared brain every AI agent you use can read and write. One structured source of truth across Claude, ChatGPT, Cursor, and coding agents.

BEIJING, China — A bipedal humanoid robot suffered a spectacular crash during a running practice session ahead of the World Humanoid Robot Games, dubbed the “Robot Olympics,” which are scheduled to begin on 22 August in Beijing, China. During a test run, the robot accelerated at full speed along a track before slamming into a […]

Schaeffler is creating its strain wave gearboxes with a forming technology, which will enable mass production of the components.

Inference Hours is back with another Kimi K3 webinar!

On Wednesday, August 26, Zain Hasan sits down with Kevin Cui and Austin Silveria from Together AI's inference team on how we actually serve K3 in production: the architecture, quality work, and optimizations behind it.

Grok 4.6 is now available via Google Enterprise Agent Platform.

There's about to be a big fight to secure access to space.

Binance's Agent OS works with tools such as ChatGPT, Claude Code, and Cursor.

Web agent economics are set by inference volume and each step is one inference call. A simple task (extract a field, fill a form, navigate a site) is 10 to 20 calls.

A multi-site workflow runs into the hundreds.

From @RaiseSummit last month: @abhshkdz, @yutori_ai

Agent frameworks increasingly package procedural knowledge as skills: instruction files an agent reads on demand, while public libraries now hold thousands of them. Which skill to read has thus become a decision the policy itself makes in the middle of an episode, yet no existing signal trains it. We show that the default remedy, outcome-rewarded RL over the candidate slate, cannot teach it, for a structural reason we identify and name selector credit starvation: under a broadcast, sequence-level advantage, the few tokens that name the chosen skill carry a vanishing share of the loss, and the credit they inherit is increasingly wrong-signed as trajectories lengthen. A correct choice is punished whenever the execution after it fails, even though the choice itself is among the most valuable decisions in the trajectory. Auditing a completed run's own training artifacts confirms all three properties, each worsening monotonically with horizon. SkillGate removes the failure by construction: it partitions the token support into two disjoint credit channels, outcome credit reaching only execution tokens, and a separate action-local advantage reaching exactly the skill-naming tokens, positive only when a trajectory's single read is the correct one. On five agentic benchmarks under a 16-candidate slate, SkillGate lifts a 9B policy from 40.8% to 53.2% trial success, well ahead of the identical budget spent on outcome reward alone, while cutting exposure to misleading candidates by two thirds and reading fewer skills.

Continuous self-improvement requires an ever-expanding pool of self-generated, diverse, adaptive goals. For language agents, existing training environment pools (hand-curated, statically synthesized, or frozen-verifier) keep the goal distribution fixed as the learner scales. We introduce SPADE (Self-Play in Adaptive Synthetic Executable Environments), a self-play RL framework in which a single LLM plays two roles: an Environment Designer that writes complete, long-horizon training environments as executable code with an OpenAI Gym-style reset()/step() interface, and a Reasoning Agent that learns to act in them. Each is a stateful, multi-turn environment (state transitions, reward functions, and verification code), so one interface spans reasoning problems and multi-step agentic tool use. The Reasoning Agent's regret is estimated using the gap between its reward with and without privileged hints; in optimizing this regret signal the Environment Designer learns to target environments at the edge of the agent's capabilities while keeping them feasible. Through extensive experimentation, we find several components critical to success: grounding the Environment Designer on documents sampled from a large pretraining corpus, and giving it an accumulated environment memory. Scaling to 30B-parameter models, SPADE improves over the strongest fixed-environment baseline by +5.3 on average across eight held-out math, science, code, and reasoning benchmarks, and lifts the tool-use setting by +5.7 on BFCL-v4 multi-turn and +13.9 on ACEBench-Agent; on the games setting, the margin over the strongest baseline grows with model scale. By making environment design itself a learnable component, SPADE takes a concrete step toward open-ended self-improvement.

A traditional scalar value head estimates value with a fixed forward pass.

TEMPO instead uses an agentic value model that can scale its reasoning and tool use at test time.

On ARC-AGI-3, TEMPO scored:

→ 31.5% above the baseline checkpoint
→ 20.6% above GRPO

DeepSeek-V4-Flash-Vision-Exp is now live on the DeepSeek API Platform! 🚀

(ATTN: ADDS photos)By Kim Eun-jung SEOUL, Aug. 21 (Yonhap) -- Galaxy Corp., a Sout

New Blog: Fast Engine Recovery: Sub-Second Engine Restart for SGLang via Weight Cache Daemon.

We built a persistent GPU process that cuts weight loading for Ling-2.6-1T FP8 from ~495s to ~0.63s, by serving post-quantized weights to new engines via CUDA IPC zero-copy.
- ~785×

At the 2026 World Robot Conference (WRC) in Beijing, Chinese company RealMan Robotics demonstrated robots performing tasks such as handling medicines, inspecting electrical panels, preparing food and setting equipment remotely. RealMan’s GLN (Global Link Network) control system allows robots to be controlled remotely with millisecond latency and collect data from the actual production process. Image […]

As we continue to push the frontier of capabilities while improving efficiency, we're dropping API and credit pricing of GPT-5.6 Sol by over 20% for the next 3 months.

The new Setup Coverage Skill detects your test setup, generates coverage in a format Codacy accepts, and adds the upload to your pipeline, across eleven language families and seven CI systems.

Ox Alpha is a reasoning model designed for coding, sustained agentic work, and production workloads. This model is free to use. 1,048,576 token context window, maximum output of 131,072 tokens.

Awesome Gemma is live on @github 🤗 A list of awesome Gemma resources, tools, and projects.

1. Model cards and collections for 16 Gemma variants
2. Setup guides for Ollama, vLLM, and LiteRT
3. Fine-tuning recipes for Unsloth, Tunix, and MLX
4. Community tutorials, apps, and

Finally a good paper testing whether agents can really post-train other agents.

(bookmark it)

They analyzed a large corpus of publicly released post-training trajectories. Across tasks, the agent locks in its training strategy at the very first step and spends the entire

Language model agents now execute bounded tasks reliably. Whether they can sustain effective decision-making over long horizons, where actions have cumulative consequences and the environment responds to their choices, remains largely unmeasured. FM-Bench (Football Management Benchmark) measures this. An LLM agent runs a football club for 20 in-game years through 26 tools and roughly 340 to 400 decision stops. It drafts a squad on the same budget as every rival, trades players, negotiates contracts, invests in facilities and youth, sets lineups, and answers to a board that can fire it, while a deterministic engine accumulates every year into one final score with no LLM judge or human rater. The solo track plays each of 15 frontier models against a frozen scripted world, and the Arena places the same models plus a scripted anchor in one shared 20-year world; to our knowledge, the first head-to-head evaluation at this scale. We measure six behavioral capabilities behind the score. Across three seeds, all 15 models complete every horizon while the blind scripted baselines die out in most of theirs, and claude-fable-5 tops the solo board on mean score and the Arena, where the title nonetheless rotates among ten models. Neither scale, price, nor vendor predicts the order; the order settles only late in the horizon, and the best first-play human lands only at the bottom of the model board. What separates the models is managerial behavior rather than computation. Higher-scoring models reduce slow-payoff investment near the end, keep cash invested rather than idle, and open renewals well before the deadline, while token spend predicts nothing. No model learns the market's hidden prices from hundreds of rejected bids, and self-managed memory fails in two opposite modes: an archive that only grows or a plan rewritten every season. Code is available at https://github.com/Analogy-AI/fm-bench.

Programmable logic controllers (PLCs) run industrial plants, and large language models can already generate independent program organization units (POUs) for them. Whether such logic integrates into an existing PLC project and then runs correctly has been checked only in limited tests. We present \textsc{SemaPLC}, a project-grounded and verification-gated agent harness assembled from conventional tools but governed by a strict completion rule. Rather than stopping when the model judges its own output adequate, \textsc{SemaPLC} declares a task complete only when logged external checks confirm it. Those checks cover the specification, the compilation, and the behavior on a live runtime. On 117 independent-POU tasks matching existing benchmarks, it attains the highest strict verified pass rate on all seven models (72.6\% mean). On a project-context track of 65 tasks whose generated logic must compile and run inside a real project, it attains the highest mean on integrated compilation, static behavior, and dynamic behavior. Of the three layers, dynamic behavior is the most revealing. We measure it by deploying the generated and the reference logic to a live PLC runtime and comparing their executed traces. All methods fall within 10 static points of one another, whereas dynamic scores separate them sharply, from 22.4 to 31.4 for the baselines against 52.2 for \textsc{SemaPLC}. Overall, our verification-gated harness raises the mean at every layer and most sharply at runtime. Execution, not static scoring, is the faithful test of whether generated control logic actually works. \textsc{SemaPLC} is open-sourced at https://github.com/midea-ai/SemaPLC.