#Agent

AI…handle messy real-world tasks, applying taste to what works in the long run, and, above all, specifying what we actually want.

AIWe'll release aggregate stats like we did in last year's MAP paper.

AICognition is making Fusion available in Devin Desktop and CLI, a harness where a frontier lead model plans and reviews while a cheaper sidekick executes. Across listed coding benchmarks, Cognition reports Fusion cuts cost per task by about 11% to 46% versus the lead model alone, while the sidekick does the implementation work. The post recommends pairing Fable 5.1 with SWE-2, and argues price per task matters more than price per token.

AIDwarkesh Patel announced a new episode featuring John Schulman, Chris O'Neill, and Beren Millidge, three AI researchers from openish companies. The discussion covers the case against recursive self-improvement, drivers of Chinese labs' progress, training of automated AI researchers, long-horizon RL, the sim-to-real gap, and the role of data and RL in recent progress.

AIJohn Schulman, Beren Millidge, and Charlie O'Neill discuss whether current training methods could produce recursive self-improvement. They argue that progress depends on whether models can learn their own objectives and on sample efficiency, and that distillation keeps frontier capabilities from centralizing quickly.

AIIt does not treat a hard question as one long search. It drafts a candidate, checks every constraint, and goes back for what is still open. 122B MoE. 10B active. On hard search benches it sits next to GPT-5.4. Here is how it works👇

AIShanghai Artificial Intelligence Laboratory has released Atria Dawn Preview, an agentic model built on the 744B-parameter MoE GLM-5.2 foundation model, with a 256K context window. The release page reports benchmark results across search, coding, tool use, productivity, and cybersecurity, and describes text-only setup for Codex and Claude Code.

Why it matters: The release page gives a full benchmark table against named rivals and setup steps for Codex and Claude Code, useful for anyone evaluating agentic models.

AIbut nvm, sometimes people can take the small unknown costs for big benefits in efficiency as fm are becoming more and more focused on so called agentic tasks

AISakana AI released Fugu Max and Fugu Ultra v2, multi-agent orchestration systems that route tasks across a pool of open-weights and specialized models. The source says Fugu Max delivers performance within striking distance of elite models at two to six times lower cost, while Fugu Ultra v2 outperforms Opus 5 and Fable 5 on Chartography and outperforms models costing three to five times more per token on DeepSWE.

AIOpenAI has launched the Agents API, a cloud-based way to build agents backed by the Codex harness. Developers can connect their favorite tools and connectors and attach agents to any sandbox. The author says the API lets firms build scaled agents and expose them inside their own internal AI applications.

Why it matters: The source describes an API for building cloud agents on the Codex harness, useful for teams planning to embed agents in internal applications.

AIGoogle Developers Blog describes autofinetune, a project applying autonomous agent loops to LLM post-training with Tunix, Gemma, and Cloud TPUs. In an SFT case study on FunctionGemma, an agent ran 20 automated experiments on a Cloud TPU v5e-1 to adjust LoRA settings, optimizers, and learning rates. In a GRPO case study on Gemma 3 1B for GSM8K math reasoning, the agent ran 40 experiments on a Cloud TPU v6e-1 and improved total reward by about 10%.

AIAmazon Science researchers propose that machine learning research agents avoid overfitting benchmarks despite years of iteration against the same tests. They attribute this to generalizable strategies being expressed compactly, leaving no room for memorization, while overfitting strategies fail to survive a compression bottleneck.

AIOpenAI has made ChatGPT for Financial Services available, a tailored ChatGPT Work experience that combines built-in financial data with GPT-6 Astra's reasoning. Teams can use it to develop research, build financial models, and create customized client materials. The author says it integrates financial data sources including Daloopa, PitchBook, and LSEG.

AIRedwood Research defines NLS depth, a measure of how much unverbalized serial computation an AI system can perform, building on Brown-Cohen et al.'s opaque serial depth. The proposal counts only nodes that output natural language initialized from a pre-training prior as interpretable bottlenecks. Standard transformers scale with their layer count, while latent reasoning architectures would raise NLS depth sharply.

AICognition has welcomed Jonathan Kelley and the Dioxus team, whose framework Cognition used extensively to build and improve Devin's performance. Cognition plans to continue supporting Dioxus, Blitz, Taffy, and Subsecond while increasing investment in Dioxus-Native and Blitz. The Dioxus team will also work on Devin's virtual machine, computer use skills, and testing capabilities.

AIAmazon Science researchers found that LLM research agents running benchmark hill-climbing rarely overfit, because their winning strategies can be compressed into prompts of about 32 tokens. A fresh reproducer agent with no access to the validation set matched the explorer's performance on most of eight datasets from that short prompt alone. The team also used the test to flag overfitting, since validation-specific gains did not survive compression.

AIBut it will never become my “intelligent personal hub” if I can’t run an agent on it that can use the same apps I can. Just like the iPad, amazing hardware is held back by an operating system that treats software as something only humans can operate. Agents using computers is the new paradigm.

AICompared with the previous generation, V4.1-Flash’s KV cache needs just: 🔹 1/4 the HBM 🔹 1/8 the SSD storage Cache-hit charges often account for a large share of agent costs. Compressing the cache cuts those costs significantly. 3/6

AIDeepSeek released DeepSeek-V4.1-Flash, a multimodal Mixture-of-Experts model with 552B backbone parameters and support for contexts up to one million tokens. The technical report says its global KV cache footprint is 890 bytes per token, roughly one quarter of DeepSeek-V4-Flash, and reports 8B activated parameters per token during prefill and 16B during decode.

Why it matters: The report shows KV cache per token falling to about one quarter of DeepSeek-V4-Flash, a concrete tradeoff between long-context serving cost and benchmark results.

AICursor is launching Projects, a beta feature for larger work such as a feature, migration, or full app, rolling out to all users starting today. A coordinator agent plans the work, delegates it to implementing agents that can run in parallel, and runs on a cloud computer so it continues when the laptop is closed. Each Project keeps shared context files synced across cloud and local machines, and subscriptions let the coordinator act on Slack channels, schedules, or PRs without a prompt.

Why it matters: The source details how a coordinator agent plans, delegates, and syncs shared context across cloud and local machines, useful for judging how long-running agent work might fit a team's workflow.

AIGenspark and Fireworks Lab post-trained the open-weight MiniMax M3 into Gen-1 Slides, a model that plans, writes, and checks slide decks end-to-end. On Genspark's evaluation it matches Claude Opus 5 at about 1/17 of its input-token list price, roughly 90% less per finished deck. In production it cut low-rated decks from 18% to 3.6% over the base model.

Why it matters: The post explains a post-training pipeline with reward design, curriculum, and numerical fixes, showing how a cheaper model was tuned toward a frontier quality bar.

AIMicrosoft Foundry's July and August 2026 updates make Hosted Agents, Voice Live integration, and Toolboxes generally available. The post adds Claude tools on Azure, Model Router region and model pool changes, Foundry Local preview features, and updated Python, JavaScript, Java, and .NET SDK versions with migration notes.

Why it matters: The roundup links each GA and preview change to code examples, migration notes, and runtime requirements, which helps developers judge what to upgrade and test first.

AITo start, Resend, Amazon OpenSearch, MongoDB Atlas, Algolia, and Clerk are now one click from the chat. Auto-setup, skills loaded. Build without leaving v0.

AIReally enjoyed collaborating with @MagnusCarlsen @SRNA_LI and Jonas Niedermüller on this.

AIMistral AI helped a European energy operator migrate 40,000 lines of Fortran 77 to C++ for a reservoir simulator with no test suite. The post explains a parity harness that checks numerical agreement between the two codebases, and a workflow where agents coder, tester, and reviewer migrate modules under human review. Its authors note the approach covered the self-contained first sprint of 40,000 of 300,000 lines and that dependent systems would bring additional challenges.

AILeave Kimi Work running on your computer and keep things moving from your phone. Work anywhere, anytime with Kimi Work.

AIAmy, a 58-year-old with no coding background, built Wally, an AI voice aid app using Manus, for her brother Jamie after cancer surgery at Brigham and Women's Hospital left him unable to speak. The app combines one-tap urgent and everyday phrases with golf news, live tournament scores, and family messages, all placed on his phone's home screen.

AIGoogle announced general availability of ADK for Kotlin 1.0, a Kotlin Multiplatform framework for building AI agents on servers and Android. Version 1.0 reaches feature parity with ADK 1.0 Core and adds Android extensions for on-device models, cloud Gemini via Firebase AI Logic, and persistent sessions and memory with Room and AppSearch. The post includes a server-side incident triage example using KSP-generated tools and skills, plus an Android financial assistant example with human confirmation for transfers.

Why it matters: The post names the new Android and server-side capabilities and the code setup, helping Kotlin developers judge whether ADK fits their agent projects.

AIFactory is now available on the Claude Marketplace, letting enterprise customers apply their committed Anthropic spend toward its autonomous software development platform. The platform automates the software development lifecycle, covering planning, implementation, testing, and security within one system, with enterprise deployment options that keep execution close to customers' code and infrastructure.

AIGoogle Developers Blog argues that teams building AI coding agents should replace end-to-end benchmark scores with behavioral evaluations that test discrete, observable actions. Examples include asking clarifying questions on underspecified prompts, running a local validator before marking a build change complete, and consulting live search for current information. The post recommends fast, deterministic unit-style checks, outcome-based LLM-as-a-judge checks for complex tasks, and batch runs that track aggregate pass rates over time.

AIAt Arm Everywhere China on September 8, 2026, Arm launched products spanning data center CPUs, mobile compute subsystems, and robotics platforms. The article says CSS for Mobile 2 integrates CPU, GPU, and neural accelerator for agent AI on phones, and Arm's Neoverse CSS N4 and AGI CPU target agent sandboxes in data centers. It also reports that Arm's Total Design ecosystem now covers over 80 partners for physical AI.

AIAmazon is giving students a free full year of Kiro, its spec-driven coding agent, at 132 universities across 18 countries. The program aims to help students move from idea to working application, the same way teams at Delta, Ericsson, Siemens, and Amazon do.

AIAn agent that gets to know you over time necessarily holds a lot of context about you. That's the source of its usefulness and the reason we built Muse to be secure, safe, and private. Read the full deep dive here:

AIMeta announced Muse, a personal AI agent designed to get things done for users across many parts of life. The product is powered by Muse Spark 1.3, and the post links to an app download and a page describing how Muse was built.

Why it matters: The announcement names Muse Spark 1.3 as the underlying model, giving readers a concrete product and model pairing to track.

AIOpenAI says a group of agents using an unreleased next-generation model produced a solution to the Navier-Stokes Millennium Prize Problem. The problem asks whether smooth three-dimensional fluid motion described by the Navier-Stokes equations can break down, and it has remained unresolved for roughly 90 years. The author, Mckay Wrigley, reposted the claim with his own remark about roughly 10k agents working in a datacenter.

Why it matters: The quoted OpenAI post makes a major mathematical claim about the Navier-Stokes problem, so readers should weigh it against the proof's verification status.

AICognition has raised over $2B at a $48B valuation, led by Andreessen Horowitz and Accel. The company says run-rate revenue grew from $492M to almost $900M since its May round, and Devin now offers Auto-Triage, Security Swarm, and Automations.