explainx / blog / topics
AI Agents
An AI agent is a model in a loop: it plans, calls tools, checks results, and keeps going until a task is done. Agent harnesses, computer use, memory, and multi-agent setups are where most of the engineering happens.
This page tracks agent releases, research, and incidents, plus our guides on building agents that hold up.
315 stories · latest Oct 29, 2026
Start here
Δτ-Bench: The Agent Benchmark Where Tools Keep Changing (Best Score 42.3%)
Δτ-Bench is a new executable benchmark that keeps user goals fixed while the tools around an agent change. 1,588 tasks across 42 domains and six ecosystem versions; the strongest evaluated system reaches 42.3% task success. Here is what it measures and how to run it.
Multiagent Systems Need Four Engineering Fixes: Infrastructure, Identity, Isolation, Kill Switches
Katelyn Lesse, who leads platform at Anthropic, wrote an essay on October 26, 2026 arguing that multiagent systems will only deliver their promise if engineers solve four problems: runtime infrastructure, agent identity and authorization, isolation, and monitoring with real kill switches. We unpack each one, add what you can do today, and note where the essay is opinion.
Prompt Caching for Agents on Amazon Bedrock: Patterns That Cut Input Cost About 62%
An AWS Machine Learning blog post published October 26, 2026 walks through prompt caching patterns for agents running OpenAI models on Amazon Bedrock. One six-turn GPT-6 Astra test saw about 62 percent lower input cost. Here are the patterns, the traps and how to measure your own savings.
Agentic Retrieval on Amazon Bedrock Managed Knowledge Bases: The AWS CDK Blueprint
AWS published a CDK walkthrough on October 26, 2026 that replaces single-shot RAG with an agent that routes across two knowledge bases, retrieves through AgentCore Gateway and returns cited answers, with dashboards and evaluation provisioned alongside. Here is what the architecture is, what it costs you to run, and where it fits.
iPhone Duo Has No Face ID: Touch ID, Passcodes and AI Agent Security Explained
Apple says iPhone Duo unlocks with Touch ID in the side button, not Face ID. This guide explains what that means for security, what Apple has and has not said, and how to set permissions for the AI assistants and agents now living on your phone.
TradingAgents: The 114K-Star Multi-Agent Trading Framework, Explained
TauricResearch/TradingAgents is trending again with about 114,000 stars and a v0.7.0 release that can run on a ChatGPT Plus or Pro plan. Here is how the analyst, researcher, trader and risk-manager design works, how to try it, and what Hacker News skeptics point out.
Timeline
October 2026
Oct 29
Bernstein: An Open-Source Governance Layer for AI Coding AgentsBernstein is a one-maintainer, Apache 2.0 tool that schedules CLI coding agents in parallel git worktrees, gates their output, and records signed receipts. Here is what it does, how to try it, what it does not do, and where it sits next to other orchestrators.
Oct 29
Stripe Pay: Send Money to Another Business From the CLI With Just a HandleOn October 27, 2026 Stripe announced stripe pay, a command that sends instant transfers to another business by its Stripe profile handle. We read the docs: it is in preview for eligible US and UK businesses, supports USD, GBP and USDC, and was built with AI agents in mind.
Oct 28
Grok Bot Vault: Saved Logins and Cards the AI Never Sees, and the No-Approval CatchGrok Bot now has a Vault for passwords, payment cards and contact details. The docs say the model never sees the secret values, but a saved card fills at checkout with no approval each time. Here is how it works, what is verified, and how to use it safely.
Oct 28
Memory Tax: Many Agent Memory Systems Lose to Plain RAG, Study FindsA new paper from the University of Illinois Urbana-Champaign compares 11 popular agent memory systems against plain chunk retrieval and finds that many score lower while costing far more. It names the gap the memory tax and shows a cheap fix.
Oct 28
Microsoft Copilot Harness in Foundry Agent Service: Preview, Skills, and Pricing From December 1On October 28, 2026 Microsoft introduced the Copilot harness in Foundry Agent Service in public preview, a managed agent loop and sandbox with skills, MCP tools and subagents, and made long-running resilience for hosted agents generally available. Here is what you get, how it differs from hosted agents, and what it costs.
Oct 27
Axiom Says Its AI Formalized the 15,000-Page Classification of Finite Simple Groups in 24 Days: Claimed vs VerifiedAxiom Math says its AI agents turned the roughly 15,000-page Classification of Finite Simple Groups into a machine-checked Lean proof in 24 days. The announcement is on X, the big numbers come from a single secondary report, and no repository or paper was visible when we looked. Here is the split.
Oct 27
Continual Learning Exam: Do AI Agents Actually Learn From Experience?A new benchmark, the Continual Learning Exam, chains five professional-task levels so an agent can only advance by reusing what it learned earlier. The best agent completes under half the levels even with a retry. Here is what it measures and how to run the public subset.
Oct 27
Δτ-Bench: The Agent Benchmark Where Tools Keep Changing (Best Score 42.3%)Oct 27
Firefox Smart Window Adds AI Form Filling and Page Watching: How It Works and What It SharesMozilla updated Smart Window, Firefox AI browsing mode, on October 27, 2026 with two features: form filling drawn from your open tabs and /watch tasks that monitor pages. The Verge reports that source page data goes to Google Gemini 3.1 Flash Lite.
Oct 27
Multiagent Systems Need Four Engineering Fixes: Infrastructure, Identity, Isolation, Kill SwitchesOct 26
android-ad-cleaner: Let a Coding Agent Switch Off Ads on Your Android PhoneA new MIT-licensed GitHub repo, android-ad-cleaner, gained about 1,500 stars in two days by handing the tedious job of switching off ads, trending-search feeds and marketing notifications on Chinese Android phones to a coding agent over adb. Here is how it works, what it will and will not touch, and what to check before you let an agent drive your phone.
Oct 26
Prompt Caching for Agents on Amazon Bedrock: Patterns That Cut Input Cost About 62%Oct 26
Agentic Retrieval on Amazon Bedrock Managed Knowledge Bases: The AWS CDK BlueprintOct 26
Jac: A Python-Like Language Built for AI Agents, ExplainedAt All Things Open in October 2026, University of Michigan professor Jason Mars presented Jac, a Python-like language that targets Python, JavaScript and native code and models data as graphs for AI agents. Here is what it is, what The Register reported, what is only a talk claim, and how to try it.
Oct 26
Mirendil Fern 0.1: A 10M-Token Context Model for Agents, What Is Published and What Is NotMirendil, the lab behind the one-researcher Mirendil-Base experiment, released its first model. Fern 0.1 supports up to 10 million tokens, but the announcement gives no benchmark numbers in text, and access is early-access only.
Oct 26
OpenBot by Nightly Labs: A Desktop Workspace for Codex, Claude Code and Grok Agentsnightly-labs/openbot gained about 312 stars in a day. It is a local-first desktop workspace where Codex, Claude Code, Grok CLI and other agents each get a workspace, a queue and a way to message each other. The README warns agents run with full access and no per-action approval.
Oct 25
iPhone Duo Has No Face ID: Touch ID, Passcodes and AI Agent Security ExplainedOct 25
LearnHarness: The Open-Source Agent Harness That Teaches You Hard ThingsLearnHarness, a new open-source desktop app by Amos Blomqvist, gained about 1,450 GitHub stars in nine days. It wraps your own model API keys in a loop of measuring, planning, teaching and practice, with a knowledge graph that persists between sessions. Here is what it does, what it costs, and where it is limited.
Oct 25
Nanbeige4.3-33B: Open Apache 2.0 Agent Model Tested Against Qwen3.8-27BNanbeige released Nanbeige4.3-33B on October 24, 2026: an Apache 2.0 dense model with a 512K context window, built as a teacher for a coming 3B model. Here is what the model card claims, which numbers are self-reported, and how to try it.
Oct 25
TradingAgents: The 114K-Star Multi-Agent Trading Framework, ExplainedOct 24
Anthropic: Fold Subagents Into Your Main Agent Loop (Skills and Tools Guide)Anthropic Applied AI says enterprise agents built as routers, planners and a subagent per team often do better as one agent loop with progressively disclosed skills and tools. Here is the argument, the migration path, and where subagents still win.
Oct 24
CowAgent: The 47k-Star Open-Source Agent Harness ExplainedCowAgent, formerly chatgpt-on-wechat, is now an open-source agent harness with three-tier memory, a skills system, MCP support, multi-agent teams and more than ten chat channels. Here is what it actually does, how to install it, what it costs to run, and where the sharp edges are.
Oct 24
humans& Boba and Ampersand v0.1: A Group-Chat AI Built on GLM 5.3 (744B)humans& launched Boba, a free text-and-group-chat assistant, and opened the Ampersand v0.1 model behind it through an OpenAI-compatible Agents API. The base is Z.ai GLM 5.3 at 744B parameters, not the 700B some headlines say, and every benchmark so far is vendor-reported.
Oct 24
NN/g Survey of 521 UX Practitioners: AI Saves Time, but 69% of Comments Are NegativeNielsen Norman Group published a survey of 521 UX practitioners on October 23, 2026. AI shortened every task people used it for, but mandates, unreliable output and inflated expectations kept enthusiasm lukewarm. Here are the numbers, the role and sector gaps, and the limits of the sample.
Oct 24
Tracely: Open-Source CI/CD That Turns Failed Agent Runs Into Regression TestsTracely is a self-hostable, MIT-licensed project that grades production agent traces, clusters failures into issues, freezes them into replayable regression cases, and blocks pull requests that would ship them again. Here is what it does, how to try it, and what to check before you trust it.
Oct 23
Atria Dawn: Shanghai AI Lab Releases a 744B Open-Weights Agent Model Under MITShanghai AI Laboratory published Atria Dawn on Hugging Face: a 744B-parameter mixture-of-experts agent model built on GLM-5.2, MIT licensed, with a 512K context window. Its card claims wins on CyberGym, WideSearch and AutomationBench, and clear losses on SWE-bench Pro. Here is how to read the table and run it.
Oct 23
Chroma Joins Replit: Chroma Cloud Winds Down in 60 Days, ChromaDB Stays Apache 2.0Chroma, the company behind the ChromaDB vector database, announced on October 23, 2026 that it signed a definitive agreement to join Replit. The managed Chroma Cloud service will be wound down over 60 days; the open-source project continues. Here is what is confirmed, what is not, and how to prepare.
Oct 23
General Reasoning RLCR: Training a 120B Model as an Agent OrganisationOn October 23, 2026 London lab General Reasoning introduced Reinforcement Learning from Collective Rewards (RLCR) and its first long-horizon model, Avalon-M1-Mini (120B). explainx.ai explains the idea, what the team says it found, and which parts still need the technical report and released weights to check.
Oct 23
Humans Are No Longer the Only Users: How AI Agents Are Rewriting the Web, Software and SecurityFor thirty years the web assumed a person on the other end. That assumption is failing: assistants shop and sign in for users, summarizer bots flood small sites, attackers run agents on stolen keys, and companies are choosing whether to ban agents or build for them. This is a detailed look at what changes for identity, payments, pricing, analytics and security, with a practical checklist.
Oct 23
MIT Left 13 Claude Agents Alone on a Virtual Island for 30 Hours. They Started Playing.An MIT paper called "Is this machine playing?" left 13 coding-agent copies on a simulated island with no goal. They stacked blocks, drew mandalas, invented sports, and later outperformed identical agents with no island time. Here is what the paper shows and what it does not.
Oct 22
1Password for Agents: Let AI Agents Sign In Without Ever Seeing Your Password1Password announced 1Password for Agents on October 22, 2026. An agent proves its identity, you approve the request on your own device, and 1Password fills the login so the agent never handles the password. Here is how it works, who supports it and the open questions.
Oct 22
Beacon: Open-Source Memory That Follows You Across Claude Code, Codex and CursorBeacon, from Asymptote Labs, hooks into more than 20 coding-agent harnesses, normalizes their session traces into one format, and distills human-approved lessons into skills any future agent can load. Here is how it works, how to install it, and what to check before you turn on cloud forwarding.
Oct 22
Claude Duet vs OpenAI Dots vs Meta Muse vs Grok Bot: Which Proactive AI Agent Should You Use?Anthropic's Claude Duet joins OpenAI's Dots, Meta's Muse and SpaceXAI's Grok Bot as an always-on agent with its own cloud computer. This comparison covers model, price, who can get it, where you talk to it, what it can do without asking, how it handles passwords, and where each has already stumbled, plus OpenClaw and Hermes for people who want to self-host.
Oct 22
Composio Firewall: Blocking Prompt Injection in Tool ResultsComposio has added a Firewall to its agent runtime. It reads tool results before your agent does, blocks hidden instructions, and lets you write natural-language rules for tool calls. The docs are specific; the security numbers are not published yet. Here is what is confirmed, what is claimed, and how to test it.
Oct 22
Databricks MemCurrent: Agent Memory That Triples Pass Rate With 31% Fewer TokensDatabricks described MemCurrent, a persistent memory architecture for production agents that now powers Agent Bricks Memory and Genie. On its internal benchmark it lifts pass rate from 22.3% to 68.4% while using fewer tokens. The headline benchmark is Databricks own.
Oct 22
MemTensor MemT-1.0: Open Models With Native Memory, 66x Smaller Than a KV CacheMemTensor open-sourced two memory-native models: MemT-1.0-Nano on Qwen3.6-35B-A3B and MemT-1.0-Flash on DeepSeek-V4-Flash. They keep a fixed-size memory state instead of replaying history, but the model card says strong text-memory systems still score higher overall.
Oct 22
Meshy Skills for Hermes Agent: Generate 3D Models and Print Files From ChatMeshy published two open-source skills, meshy-3d-generation and meshy-3d-printing, for Hermes Agent. They drive the Meshy CLI to create models, textures and animations and to prepare prints. Model generation needs a paid Meshy plan.
Oct 22
Flint-interactive: Microsoft Research Gives AI Agents Clickable, Linked ChartsMicrosoft Research extended its Flint chart language with Flint-interactive, so an agent can describe brushing, highlighting and cross-chart links in a short spec while a compiler handles the rendering logic. It runs in chat through Flint-MCP and MCP Apps.
Oct 22
Survey of 1,700 Managers: AI Makes Employee Work Harder to Judge, and Review Time Is RisingOct 22
Sage-A1.5: Shanghai AI Lab Open 35B MoE with Subagents and Verification ModeOct 22
Spotify Bets AI Agents Are a Tailwind: What Söderström Said and What Is UnprovenOct 22
Walmart Agent Gateway: An MCP Server and Agent Profiles So AI Agents Can Shop With LimitsOct 21
OpenVLHarness: UCLA's Visual Agent Harness Adds 10 Points Without Fine-TuningOct 21
S-Agent 2.0: Give a Coding Agent a 3D Perception CLI and It Beats SpatialClawOct 20
Exa 3.0: What the Search API Claims on Eight Benchmarks and What You Can CheckOct 20
ALTK-Evolve: IBM Research Lifts Agents on tau2-bench by Learning Rules From Their Own RunsOct 19
Reverse Engineering With an AI Agent: How to Prove It WorkedOct 19
Stripe Link Says Agent Purchases Grew 38x in September: What Is Claimed and What Is VerifiedOct 19
World ID for Banks: Proof of Human, Zoom and AI Agents, Claimed vs VerifiedOct 18
Taste-Bench: Even the Best AI Agent Picks the Right Path Only 59.7% of the TimeOct 17
Citi Joins JPMorgan in Blocking AI Agents From Customer Accounts: What Is Confirmed and What Is NotOct 17
India Is the Latest Country to Report Rogue AI Agents, and This Time It Was Several ProvidersOct 17
Orca (stablyai/orca): The Open-Source Agent Orchestrator With Worktrees, Native Chat and a Phone CompanionOct 17
Solid Glass: An Open-Source Design System for AI Agent Interfaces, Built on shadcn/uiOct 17
Workday to Acquire C1.ai (ConductorOne): Why HR Wants to Be the Off Switch for AI AgentsOct 16
OpenGCUA: Alibaba Opens the Stack for Generalist Computer-Use AgentsOct 16
CERT-In Warns AI Agents Read Indian Organisations' Confidential Files UndetectedOct 16
decomp.dev Is Shutting Its Index Because AI Made Game Decompilation Too EasyOct 16
ESP32 iPhone Agent: A $7 Board Lets an AI Model Use a Real iPhoneOct 16
FDE-Gym: A Benchmark for the Last Mile of AI Delivery, Where the Best Agent Resolves 14 of 95 TasksOct 16
Tencent Hy-Memory-30B-A3B: An Open 3B-Active Model That Writes Agent MemoryOct 15
When an AI Interviewer Meets an AI Interviewee: The Bots Booked a Second Round Without HimOct 15
Capsem: An Open-Source VM Sandbox for Claude Code, Gemini CLI and CodexOct 15
Figma Specs: A Markdown Rulebook for AI Agents, Plus Code Layers in Public BetaOct 15
How to Build an Always-On AI Agent: Architecture, Triggers, Memory and DurabilityOct 15
Rivian Says Third-Party AI Agents Like Claude Are Coming to Its CarsOct 14
Agent-Reflex: Training Qwen3-8B Agents to Recover From Broken Tools and User CorrectionsOct 14
CLI-Anything: Turn Any Software Into an Agent-Ready CLIOct 14
Expand Fetch: An Agent Web-Reading API Claims 74% on a Hard BenchmarkOct 14
Grep AI Launch Week: Skills Library, Decision Outputs With Jev, and a 1,000-Integrations ClaimOct 14
Hermes Agent v0.22.0 "Threshold": Free Onboarding, One-Click Local Models and a Stable ChannelOct 14
How AI Agents Actually Decompile Games: The Snowboard Kids 2 Workflow, Its Limits and What Is VerifiedOct 14
HybridAgent: Together and Stanford Let Agents Think in Latent Space, Act in TextOct 14
iFixAi: An Open-Source Audit That Grades Your AI Agent A to FOct 14
McKinsey Adds an AI Avatar Client to Its Job InterviewsOct 14
Scion vs Git vs Cursor Origin: Three Ways to Handle Code at Agent ScaleOct 14
Server-Centric vs Distributed Version Control for AI Agents: A Decision GuideOct 14
A Solo Founder Let Grok Bot Run PromptBase for a Week: What He ReportedOct 14
Swarmchasers Track a Chinese AI Agent Fleet Hitting Alibaba Amap From Tencent CloudOct 14
Top 10 Gemini 4 Argon Use Cases: CAD Animation, Mars Data, Pagodas and Real Google WinsOct 13
Shopify Opens Its Catalog to Any AI Agent: What the Shop Rules SayOct 13
Tesla's Digital Optimus Plays Halfway Through Diablo: What Musk Claimed and What Is VerifiedOct 12
DigitalOcean Signals and Simulations: Catching AI Agents That Fail SilentlyOct 12
Fork-and-Flush: Why Clearing Chat History Helps Long-Running AI AgentsOct 12
SubQ Snap: Subquadratic's Decision Model Reads 384K Tokens for $0.05 per MillionOct 11
AI Agent Teams Cost Up to 5.1x More Than One Agent for Barely Better ResultsOct 10
AI Agent Privacy Promises: Meta Muse vs OpenAI Dots, and What Holds UpOct 10
Pine Computer Scores 78.3% on SaaS-Bench, but Finishes Fewer Tasks: Claimed vs VerifiedOct 10
Talorys: A Personal AI Agent You Deploy to Your Own Cloudflare Account in One CommandOct 9
Agent Swarm by Desplega: Open-Source "Company OS" for Claude Code and Codex WorkersOct 9
Firecrawl Universal Scrape: One /scrape Call for Web Pages, Files and 200+ Data ProvidersOct 9
Prime Agent Rust Port Merged: What Is Claimed vs. VerifiedOct 9
Reverse Engineering Is Now a Prompt: What REA's Case Studies ShowOct 8
Paul Graham: Amazon Banning AI Agents Is a Startup OpportunityOct 7
Et Tu, Brute? Study Finds AI Agents Steer Wealthier Users to Pricier Flights, Insurance and SchoolsOct 7
Agents Without APIs: Gumloop Agent Browsers and AgentMail's AgentID ExplainedOct 7
Hark Pro Explained: Brett Adcock's Proactive AI Assistant vs Muse, Dots and Grok BotOct 6
Agent Teams Are the New Org Chart. The Web Is Starting to Lock the Door.Oct 6
Gamma 5: An Agent-Driven Rebuild Aimed at Killing the "AI Smell" in SlidesOct 6
Personal Agent Protocol: Sierra, Meta, Walmart and Stripe Back a New Open StandardOct 4
Agents Don't Need Memory, They Need DocumentationOct 4
Vercel Confirms a KVM Zero-Day VM Escape: What It Means for Agent SandboxesOct 3
Cua Spaces: Cross-Computer AI Agents on Your DesktopsOct 3
Is the Harness the Company? What Shrivu Shankar Gets Right — and Where He OverreachesOct 3
Perplexity Computer Built a Map of 25,868 NYC Restaurants: Where the Data Came From and What It ShowsOct 2
Pi 1.0 and Pi Durable: How People Actually Use a Minimal HarnessOct 1
LiteLLM Lens Puts Agent Debugging on the Gateway You Already Run
September 2026
Sep 29
Know Your Agent (KYA): Baselayer Identity Suite for BanksSep 29
ElevenLabs Eleven v4 and v4 Turbo: Expressive TTS Ranked #1 by Artificial AnalysisSep 28
Hindsight: The Open-Source Memory System That Makes Agents Learn, Not Just RecallSep 28
Manus 2.0: Cascade Harness, Manus Studio, Cloud Computer, and Cue AgentsSep 28
Paperclip: The Open-Source App for Running a Company Made of AI AgentsSep 27
AI Roll-Ups: How a Services Back Office Runs on AgentsSep 25
OpenClaw Deleted 400K Lines of AI-Written Tests. The test-audit Skill Shows HowSep 24
Contrastive Language Model (CLM-8B): An Open System One Model That Matches Jev at Up to 9x Lower LatencySep 23
DigitalOcean Launches Managed Agents: Pay-Per-Use CPU and 16,000 ToolsSep 23
Do Subagents Actually Use More Usage? Yes — Here's Why, With Real NumbersSep 23
Stripe Upgraded 7.8 Million Checkout Pages to Cut AI Token Use 42%Sep 23
Thariq's Case Against Shipping 10x More Features Just Because Models Got SmarterSep 23
Unreal Agent: An Async Tool-Calling Harness That Cuts Coding-Agent Costs 40%Sep 22
AgentRun: How Grep.ai Turns Agents Into Cheap, Auditable WorkflowsSep 19
Agentic AI Certifications in 2026: Which Ones Are Actually Worth ItSep 19
What Is an 'AI Agent Workforce'? Running Multiple Agents Like a TeamSep 19
AI Evals, Explained: What Engineers and PMs Actually Need to BuildSep 18
Exa Snapshot: Search the Web as It Looked on Any Past DateSep 18
Google Labs Launches CC, an AI Agent for Family LogisticsSep 18
Top 10 Harness Engineering Concepts Every AI Builder Should KnowSep 18
What Is Harness Engineering? The Layer That Turns a Model Into an AgentSep 17
Monid Open-Sources a Tool Router With 2,000 APIs for AI AgentsSep 16
Elon Musk Cited "The Machine Stops." Here's the Real Lesson for AI Teams.Sep 16
Plasma AI Radio: A Shared Chat Channel for Cross-Provider AgentsSep 16
Tencent Open-Sources BrowserSkill: Agents Borrow Your Real, Logged-In BrowserSep 15
AI Agents Breached 395 Organizations Through PaperCutSep 15
Pion: The AI Agent Andon Labs Built to Run a Company AutonomouslySep 15
Put Claude Fable 5.1 and Gemini 3.8 Flash on the Same Coding TeamSep 15
MDM Isn't Enough Anymore: Managing AI Agents on Company DevicesSep 13
Is AGI Already Here as a Swarm? The Multi-Agent Emergence ArgumentSep 13
What Is Soul.md? Meta Muse's AI Persona File, ExplainedSep 13
Is "Harness" Software the Only Startup Left? The YC Batch DebateSep 11
How AI Agents Actually Edit Code Using PythonSep 11
Hermes Agent Manual Subagent Control: Steer, Stop, and Watch LiveSep 11
Design Docs Are All You Need: SMART Turns Specs Into Regenerable ML Perf CodeSep 10
The Second Writer: How Self-Evolving Coding Agents Actually LearnSep 9
How to Actually Work With AI Agents: A Practical Communication GuideSep 8
Supermemory Launches Learner-1 for Continual Agent LearningSep 7
Jay Alammar Publishes a Build-From-Scratch AI Agent Guide (300 Figures)Sep 6
AI Agents Keep Building Simulations Inside Their SimulationsSep 5
AI Resolves Your Incidents. Are You Still Able To?Sep 5
Solana Payment Channels: 1M Payments/Sec Claim for AI Agents, ExplainedSep 4
HarnessDev: Can LLMs Build and Evolve Their Own Agent Harness?Sep 3
fable51-worlds: An Agent Swarm Rebuilt Union Square in 3D for $33Sep 3
Context Caching in Agent Harnesses: Google's Numbers, and the Ones It Left OutSep 1
Celeris-1 Magnus Ships — Agentic Model Built for Tool LoopsSep 1
Coinbase Agents Can Now Trade Stocks — AiFi Goes Beyond CryptoSep 1
Garry Tan Ships GBrain Evals — But Who Grades the Grader?Sep 1
Manus Resumes Independent Operations: What Changed After the Meta SplitSep 1
Vercel DESIGN.md: Spec-Driven UI That Fights AI Slop
August 2026
Aug 31
Ethan Mollick: Agency and Agents — Twilight Factory vs Dark FactoryAug 31
OpenClaw 2.0 (v2026.8.1): 933 Contributors, Rebuilt Browser UI, Shared Cloud SessionsAug 30
EvoHarness-RL: An 8B Model Matches Claude Opus 4.5 on ALFWorldAug 29
Agentic Commerce Goes Live: Stripe Link for Agents and Grok ShoppingAug 29
Firecrawl Relaunches Free Keyless Search and Scrape for AI AgentsAug 29
Perplexity Search API Scores 80 in Index Debut, Ahead of Parallel and BraveAug 29
Sapient PRAXIST vs Claude Opus 4.8 on MLE-Bench: What the Number MeansAug 29
How Uber Runs Coding Agents Cost-Effectively at ScaleAug 28
PRAXIST Beta: Sapient Intelligence Open-Sources Cumulative Research AgentsAug 27
OpenExecutive: The Open Source AI CEO Built by Laid-Off DevelopersAug 26
NVIDIA NemoClaw CVE-2026-65105: One Webpage Can Poison Local OllamaAug 26
Warmwind OS 1.0: The First AI Operating System — or a Cloud Employee?Aug 25
Apodex 1.1: Asynchronous Agent Team Arrives, Plus Open-Source FrontierAgentAug 25
Garry Tan: Systems of Record Must Become AI Harnesses — or Get ReplacedAug 22
Barehands: Give AI Agents Hands With a WebcamAug 22
What Is a Gauntlet Loop? The Builder-Critic Prompt Pattern ExplainedAug 22
Instinct AI Kept Emails After Access Was Revoked — The Real LessonAug 21
MiniMax Design Is Live: An Agent That Orchestrates GPT Image 2, H3 & MoreAug 20
Block Berd: An Open-Source Desktop Home for Goose, Claude Code, and CodexAug 20
Claude Platform GA: Computer Use, Browser Tool, Skills API, Files APIAug 20
OpenBot: CopilotKit's Open-Source AI Coworkers With a Computer EachAug 20
How to Run Loops in GitHub Copilot: VS Code Agent Mode and Copilot CLIAug 20
How to Run Loops in Microsoft 365 Copilot (Honest Limits)Aug 19
LangSmith Tuned Evaluators: Perceived Error at 82% Lower CostAug 18
Cordis and Spatiotemporal Composability, ExplainedAug 18
macOS Harness: browser-use's Open-Source Tool Gives Agents Six Raw Primitives, Not App-Specific ToolsAug 18
Nous Research Ships Bot Mode: Multi-Agent Teams in Hermes DesktopAug 17
Loop Engineering: A Global Career Guide for StudentsAug 17
Loop Engineering for Students: Career Guide, Degrees, and How to Actually Build OneAug 15
OpenRouter Web Search Benchmarks: How to Pick a Search Tool for AgentsAug 14
A Real Claude Code Loop Orchestrator: Heartbeats, Tickets, and Silent BugsAug 12
DoorDash Flux: Cloud Agents, 130k Tasks, 25k Reviews a WeekAug 12
Top 5 Loop Engineering Courses in 2026Aug 11
ChatGPT Books Your Table Now: Inside the OpenTable, Resy, and Yelp IntegrationsAug 11
"Humanising LLM Outputs Is Dumb" — The Case for Rendering at the BoundaryAug 11
Manus Splits From Meta: What Gets Deleted, and How to Back Up Before August 23Aug 10
AI and Bot Traffic Overtook Humans — What Builders Should Actually DoAug 10
OpenClaw Cancelled a Stranger's Gym Booking — Australia's First "Autonomous Cyberattack"Aug 10
Spotify Xirp: A Vendor-Neutral Environment for AI Coding AgentsAug 8
Prime Agent: Prime Intellect's Self-Improving RLM Coding AgentAug 6
Humans Missed 1 in 3 AI Agent Threats: Alex Wauters's 40,000-Play DataAug 5
From ReAct Loop to Production Harness: DAG Planning, Tiered Memory, Budget PressureAug 5
LoopX: A Control Plane for Long-Running AI Agent WorkAug 5
Pokee-Isaac 28B: A Real 10M-Token Context Model on One GPUAug 5
Why AI Agents Haven't Gone Mainstream (Yet)Aug 3
Comp AI Open-Sourced an Agentic CRM — Agent First, Database SecondAug 3
TencentDB Agent Memory v2: Team Hub for Chat, Skills, Wiki, CodeGraphAug 2
Fable-OS Is a Real Kernel—But Not Yet a Bare-Metal AI ComputerAug 1
Flint: Microsoft's Chart Spec for AI Agents, Explained
July 2026
Jul 26
What a $600-a-Day AI Agent Workflow Looks LikeJul 26
What Running an AI Agent Actually Costs Per MonthJul 26
Context Window Pricing, DecodedJul 26
How AI Agents Actually Work, End to EndJul 25
Claude 5 Context Engineering: Stop Over-Constraining the ModelJul 24
Block Buzz: Self-Hosted Nostr Workspace for Humans and AgentsJul 22
Jack Dorsey's Buzz: Team Chat, AI Agents, and Git Hosting in One Nostr-Signed WorkspaceJul 22
Markdown for Agents: What HTML-to-Markdown Content Negotiation Actually DoesJul 22
Top 10 Closed-Source and Open-Source Agent Harnesses (2026)Jul 22
TryAI Canvas Arena: GPT-5.6 Sol Beats Claude Fable 5 at Drawing — for 1/20th the CostJul 21
Graphs vs. Loops: The Agentic AI Orchestration Debate, ExplainedJul 21
jcode Agent Harness: Swarm, Memory Graph, and Multi-Session RAM EfficiencyJul 20
Bojie Li's AI Agent Book: Open-Source Textbook, 10 Chapters, and Runnable CodeJul 18
Graph Engineering: After Loops, This Is How You Wire Multi-Agent Orgs (2026)Jul 17
TryAI $100 Music Video Arena: Fable 5 vs GPT-5.6 Sol Autonomous Video AgentsJul 16
DoorDash dd-cli: Order Food From Your AI Agent in the TerminalJul 14
OpenClaw v2026.7.1 — 532 Contributors, Control UI Overhaul, GPT-5.6 & Muse SparkJul 12
JPMorgan AI Agents Beat 60/40 in 20-Year Backtests — What the Numbers MeanJul 9
OpenClaw Foundation: 501(c)(3) Non-Profit, Full-Time Team, Major PartnersJul 3
Azure AI Apps and Agents Developer (AI-103): what the exam tests and how to prepareJul 3
Microsoft Foundry naming explained: Azure AI Studio → Azure AI Foundry → Microsoft Foundry
June 2026
Jun 30
Andrew Ng's Three Loops for Building 0-to-1 Products with AI AgentsJun 30
OpenClaw iOS and Android Apps Launch: Agents in Your Pocket (June 30)Jun 29
Context vs Prompt vs Loop vs Harness Engineering: The Four-Layer Agent StackJun 29
Error Propagation in Multi-Agent Systems: Structured Context Over Generic FailuresJun 29
Types of AI Agents: Complete Taxonomy and When to Use Each (2026)Jun 28
Agentic context design: how to engineer the context window for multi-turn AI systems in 2026Jun 28
How to Build an AI Agent Loop: Triggers, Retries, Checkpoints, and Human HandoffsJun 28
Context engineering vs prompt engineering: a precise distinction for 2026Jun 28
Conversation history management for AI agents: what to keep, compress, and drop in 2026Jun 28
Human-in-the-Loop AI: When to Let the Agent Run and When to Stop It (2026)Jun 28
RAG and context injection: designing retrieval pipelines that actually work in 2026Jun 28
Token budget planning and execution: how to manage context costs in production AI systems in 2026Jun 28
Tool definition and schema design: the context engineering layer most teams get wrong in 2026Jun 27
Context engineering: the complete guide to designing what your AI model actually sees in 2026Jun 27
Multi-Agent Orchestration Patterns: A Production Guide (2026)Jun 27
What Are AI Agents? A Plain-English Beginner's Guide (2026)Jun 25
Higgsfield Supercomputer 2.0: Autonomous Marketing Agent on NVIDIA (2026)Jun 24
Hermes Agent vs OpenClaw: Which Open-Source AI Agent Should You Use in 2026?Jun 24
Qwen-AgentWorld: The First Language World Model for General AI Agents (2026)Jun 24
Top 10 Things You Can Do With Hermes Agent in 2026Jun 24
Top 10 Things You Can Do With OpenClaw in 2026Jun 24
Top 25 OpenClaw Claws Worth Installing in 2026 — Ranked by Usefulness and DownloadsJun 24
What Is OpenClaw? The Open-Source Personal AI Assistant You Run On Your Own DevicesJun 23
Eric Xing Critique of Agent Model: Agentic vs Agentive AI and the GIC ArchitectureJun 23
1,009 Tokens Per Second: Mercury 2 and What Diffusion LLMs Change for Agent LoopsJun 23
Prompt Caching: Decision Framework for LLM Cost, Latency, and Security (2026)Jun 23
Stripe Directory: The Infrastructure Layer That Makes AI Agent Commerce RealJun 22
Sakana Fugu: One Model API to Orchestrate All the OthersJun 21
The Knowledge Worker's Guide to AI Agents in 2026Jun 20
AI for Business Leaders: What Actually Matters in 2026Jun 20
How to Build Your First Agent Loop: A Step-by-Step Guide (2026)Jun 19
Matthew Berman Loop Library: Free Agent Workflows for Developers (2026)Jun 19
Pi Agent Harness: Mario Zechner's Minimal Coding Agent You Can Own (2026)Jun 19
Top 10 AI Agent Loops for Coding Workflows (2026 Guide)Jun 17
Vercel eve: The Open-Source Agent Framework That Does for Agents What Next.js Did for the WebJun 17
What Is Self-Harness? The AI Agent Pattern That Improves Its Own ScaffoldingJun 16
AI Agents That Play GeoGuessr — Browser Use v4 and the Rise of Visual Geolocation AIJun 16
Loop Engineering Is Now the Most-Discussed AI Skill on Developer TwitterJun 16
What Are AI Agents? The Complete Explainer for 2026Jun 16
What Is an Agent Harness? The Scaffolding Layer That Makes AI Agents ReliableJun 14
Karpathy LLM Wiki: The Pattern Behind Agent Memory (Complete Guide)Jun 14
OKF Sample Bundles: GA4 E-commerce & Bitcoin BigQuery Datasets GuideJun 13
What Is Loop Engineering? The New Paradigm Beyond Prompt EngineeringJun 12
Mastercard Agent Pay for Machines (AP4M): AI Agent Payments ExplainedJun 10
Self-Harness: AI Agents That Improve Their Own Operating FrameworkJun 9
Anthropic VirBench: Why Biological Agents Need Deterministic Tools Like gget virus (2026)Jun 9
Build AI Marketing Agents with Claude: Complete Tutorial【2026】Jun 9
Loop Engineering: How to Design Coding Agent Loops That Run While You Sleep (2026 Guide)Jun 6
AI Agents for Influencer Marketing: How Automation & Fake Detection Are Reshaping the Creator Economy in 2026Jun 4
Microsoft SkillOpt: Self-Improving Agent Skills Guide 2026Jun 1
Hermes WebUI: The Self-Hosted AI Agent Interface That Remembers Everything (2026 Complete Guide)
May 2026
May 31
What Are Monorepos? A 101 Guide for JavaScript, Python, Next.js, AI Agents, and MCP (2026)May 30
Is OpenClaw Safe? The Complete Story of Anthropic's Ban, Peter Steinberger's Suspension, and What Users Need to KnowMay 30
Microsoft SkillOpt: The Self-Evolving Agent That Trains Documents, Not Models (52/52 Wins)May 23
Agent Markdown Files: The Complete Guide to SKILL.md, AGENT.md, CLAUDE.md, and MoreMay 22
Dotnet Skills: The Official Microsoft Repository for AI Coding AgentsMay 22
Google Search I/O 2026: The Rise of Search Agents and Agentic CodingMay 22
Understand Anything: Turn Any Codebase into an Interactive Knowledge GraphMay 20
Agency Agents: 144+ AI Specialists to Transform Your Workflow in 2026May 20
The Agentic Era: How AI Agents Will Transform Everything (2026-2030)May 17
Agent Skills: The Secure, Validated Registry for Professional AI Coding AgentsMay 14
Android 17, Gemini Intelligence, and Google Books: The 5,000+ Word Definitive Encyclopedia of the 2026 Google OS RevolutionMay 14
Higgsfield AI Supercomputer: Building a Cloud-Native Architecture for Autonomous Media ProductionMay 13
AI Native Economics: The $600/Day Agent vs. The $20 Meal LimitMay 12
Goal mode for AI agents: what it is, how to use it, and why OpenClaw, Hermes, and Codex are all adopting it in 2026May 9
Hermes Agent Hits #1 on OpenRouter Global Rankings — What 271 Billion Tokens Tells UsMay 8
DESIGN.md Templates: The Professional UI Blueprint for AI AgentsMay 8
Top 10 DESIGN.md Registries & Templates Directories (2026)May 8
What is MEMORY.md? The Long-Term Brain for AI AgentsMay 7
ByteDance DeerFlow 2.0: Open-source super agent harness with skills, sub-agents, and sandboxesMay 7
RAG vs Agentic RAG: why search beats embeddings for code retrievalMay 6
CocoIndex: incremental indexing for always-fresh agent and RAG contextMay 4
Agent harness engineering: when the model stays fixed and the scaffolding winsMay 4
Cofounder 2: superoptimizer orchestration for a multi-agent companyMay 2
OpenClaw meets ChatGPT Plus: OpenAI’s subscription path vs Claude limits
April 2026
Apr 24
DESIGN.md: the open spec that teaches AI design intent, not just tokensApr 23
Gibberlink and the “secret AI language” moment: ggwave, hackathons, and what is actually going onApr 22
What is Hermes Agent, and how does it work?Apr 13
holaOS (Holaboss): an open agent environment for workspaces, memory, and long runs