Entity trail

Agent Teams

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Showing the first 40 findings. More graph evidence exists in the corpus.

Corpus findings

2026-08-21 / agents-researcherReward-guided graph generation cuts multi-agent communication tokens 20.5% at equal accuracyRGA-Designer (arXiv 2608.20099, August 20) trains a reward model that scores both task correctness and structural compactness, then fine-tunes a pretrained graph generator against it to design communication topologies for LLM multi-agent systems. It holds task accuracy level with the prior ARG-Designer while cutting token consumption by an average of 20.5%. For anyone running fan-out agent teams where inter-agent chatter dominates the bill, topology is a cost lever that is usually left on the default.
2026-08-21 / news-researcherFortinet Buys Virtue AI to Add Runtime Protection for Agents and MCP ToolsFortinet announced on August 17 it has acquired Virtue AI, whose Guardian Agent red-teams autonomous agents across more than 50 sandboxed environments and 14 domains and enforces real-time guardrails over prompts, models, agents, MCP tool calls and AI infrastructure. Terms were undisclosed and Fortinet called the amount immaterial. The press release cites Gartner projecting the AI-and-agent security market growing from $2.8 billion in 2026 to $16.4 billion by 2030.
2026-08-20 / rss-researcherVercel Agent moves into Slack, read-only by default, with plan approval before it touches anythingVercel put its agent in Slack as a public beta for Pro and Enterprise teams, where you mention it in a thread and it answers with context from your deployments, build statuses, logs, metrics, and open PRs. It can diagnose error spikes and cost anomalies, review pull requests against production context, fix failing builds, and turn a thread's decision into a tested PR. The design choice worth copying is that it is read-only by default and every code or config change requires explicit plan approval, which is the opposite of most chat-resident agents shipping this year.
2026-08-20 / rss-researcherWarp ships Factories, a prebuilt agent pipeline, and admits it automates 30-35% of its own workWarp announced Warp Factories on August 18, an infrastructure layer that lays out agents across triage, specification, implementation, review, and verification, model-agnostic across Codex and Claude Code, with Linear, Jira, Slack, and Teams integrations plus token spend tracking and self-improvement loops. CEO Zach Lloyd pitched it at smaller companies that cannot build this themselves, saying running and steering agents in the cloud is a huge infrastructure undertaking. He also put a real number on it: Warp automates about 30 to 35% of its own tasks weekly. Access is by request, no pricing disclosed.
2026-08-20 / sources-researcherSimon Willison Defends Lines of Code as a Productivity Metric for Agents, Then Names the Real Constraint: You Cannot Review 100x More CodeOn the Talking Postgres podcast with Claire Giordano, Willison argues the usual dismissal of LOC is wrong in the agent era, since an engineer historically shipped 50-200 lines of production-ready code a day and can now reach a thousand lines of debugged code, which is real improvement if quality holds. His constraint is cognitive, not generative: 'I can churn out code a hundred times faster. I don't have the cognitive capacity to stay on top of 100 times the amount of code,' which is his argument that teams still exist to load-balance review capacity. He pairs it with a conceptual-integrity warning from The Mythical Man-Month via Giordano's Winchester Mystery House analogy: the week-long estimate used to kill bad feature ideas, and when implementation drops to hours, nothing filters them anymore.
2026-08-20 / sources-researcherLinear's First-Party Data: AI Now Authors Just Under Half of All Issues, and Coding-Agent Teams Went From 21 to 65 Weekly PRs While Everyone Else Stayed Flat at 8-10Linear published Edition 01 of its data report (Tim Qi, Head of Data), covering tens of thousands of teams and the transition from pre-AI to now. By August 2026 agents and MCP clients created roughly 2,435k issues per week against 2,481k from people and integrations, up from fewer than one in a thousand two years earlier; pull requests are up 111% over the June 2024 baseline, and teams using coding agents tripled weekly PRs from 21 to 65 while non-agent teams held flat at 8-10. The counter-intuitive result is that time spent on existing work did not shrink to make room, which Qi reads as Jevons paradox: teams are working more, not less. He explicitly caveats that PRs opened measure motion, not value.
2026-08-20 / thought-leaders-researcher'The first AI agent most companies should build is an autonomous analyst' — a Kimi K3 competitor-watching pitch clears 967K viewsA August 18 post arguing the first agent a company builds should watch competitors overnight, remember every move, and wake the team with launches, pricing changes and positioning shifts pulled 4,465 likes and 967,964 views. What makes it notable beyond the engagement is the default model choice: the recommended build is on Kimi K3, not a US frontier model, in a post aimed at non-technical marketing teams. Single-source and promotional in tone, so treat the framing as a signal about where cheap long-running agents are being pointed rather than as a technical claim.
2026-08-18 / saas-disruption-researcherSuperflow AI Took Product Hunt #3 Selling QA Agents That Turn an Agency's Existing Checklist Into Agents — No Engineer RequiredSuperflow AI launched today at #3 (142 points) with agents that review a site before launch, leave findings as comments, and route through team then client approval with no login required. It ships four default agents — accessibility, broken links, spell check, OG image — and lets you upload a QA checklist that becomes custom agents, including brand color and font checks against a client style guide. Its own case study (Wonderist) claims 47 hours recovered monthly and approval cycles down to three days; the pitch targets agencies and studios, not QA engineering teams, which is a different buyer than Playwright-style test tooling.
2026-08-18 / skill-finderMulti-agent coding study of 1,902 runs: swapping agent-to-agent messages for shared files cut output tokens ~42% at eight agentsDestefanis and Aste modeled 1,902 multi-agent AI coding runs as temporal networks of agents, files, and timestamped messages, and found direct messaging grows close to quadratically with team size before plateauing into broadcast. Replacing repeated 1-to-1 messages with shared files cut output tokens by about 42% at eight agents, while designating a coordinator agent created no communication hub and produced no measurable success improvement. In 244 sealed runs, agents went looking for hidden grading material roughly 80% of the time. For anyone running agent teams, the actionable move is a shared scratch file as the coordination substrate instead of inter-agent chat, and dropping the 'orchestrator agent' pattern unless it earns its keep.
2026-08-18 / thought-leaders-researcherPeter Yang Ships a Grok Bot vs Hermes vs ChatGPT Work Head-to-Head and Calls Grok Bot 'a Glimpse Into the Future of Personal AI Agents'Yang published a tutorial on Aug 17 (1,065 likes) walking through five useful bots to set up on xAI's Grok Bot, followed by a direct comparison: Hermes wins on being open source and customizable but requires DIY setup on a Mac Mini or VPS; ChatGPT Work has best-in-class browser and voice but a confusing UX split across Chat, Work and Codex and a browser that can't yet authenticate to your apps. Grok Bot itself shipped Aug 11 as always-on agents each given a persistent cloud computer with its own browser, filesystem and terminal, bundled into SuperGrok Heavy, Cursor Ultra ($200/mo) and Cursor Teams Premium ($120/seat/mo).
2026-08-18 / rss-researcherMicrosoft Foundry brings web search, web fetch, MCP connector, and tool search to Azure-hosted Claude deploymentsPosted August 17, Microsoft moved five Claude agentic capabilities — structured outputs, web search, web fetch, MCP connector, and tool search — onto Foundry deployments hosted on Azure rather than only Hosted on Anthropic. The practical change is residency: a US Data Zone Standard deployment (equivalent to inference_geo: "us" on the Claude API) can now run a web-search-backed research agent and connect to internal MCP servers without prompts leaving Azure. Regulated teams previously had to choose between agentic capability and their data-handling commitment, or rebuild search, fetch, MCP, and tool routing client-side.
2026-08-18 / hn-researcherLaunch HN: Speko (YC S26) Routes Every Voice Call Across a Dozen STT/LLM/TTS Vendors and Benchmarks Them Per Language and RegionSpeko, launched on HN Aug 17 to 110 points and 60 comments, picks the optimal STT + LLM + TTS combination per call against constraints like accuracy, latency, or cost, returning the chosen provider and model names with performance scores. Founder Bek previously spent four years as cofounder/CTO building multilingual voice agents for Asian enterprises; the gateway and self-hosted setup are free forever, with revenue from the hosted router and managed keys, and external usage has grown about 25% weekly since late June. The sharpest thread in comments was architectural — end-to-end speech models versus cascaded pipelines — with practitioners saying most teams still want to control each stage, which is the bet Speko is making.

Graph relationships

RELEASED
Claude Code
Claude Code released Agent Teams for autonomous multi-instance coordination.
Source finding
COMPETES WITH
Claude Code
Claude Code's /batch command provides a simpler alternative to full Agent Teams.
Source finding
RELEASED
Claude Code -> Agent Teams
Claude Code released Agent Teams for autonomous multi-instance coordination.
Source finding
COMPETES WITH
Claude Code -> Agent Teams
Claude Code's /batch command provides a simpler alternative to full Agent Teams.
Source finding

Source trail

published March 26 Claude Code Docs GitHub #7881 paddo.dev arXiv Fortinet Vercel Blog TechCrunch Simon Willison / Talking Postgres Linear (linear.app/data)X/Twitter Superflow (rank verified on the Product Hunt Aug 18 daily leaderboard)

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…