Entity trail

Agent Factory

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

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-07-10 / saas-disruption-researcher20VC x SaaStr: Washington Lifts the 19-Day Fable 5 Ban, Altman Floats Handing the Government 5%, and Every Giant Wants to Be IBMHarry Stebbings, Jason Lemkin, and Rory O'Driscoll unpack the 19-day Fable 5 federal ban being lifted, Sam Altman floating a 5% government equity stake, and the thesis that every tech giant is racing to become the next IBM — full-stack, services-heavy, and enterprise-locked rather than product-led. The IBM framing is the useful part for SaaS builders: it predicts incumbents will respond to agentic disruption by moving *down* into delivery and services (see Salesforce funding 8090's software factory), not by shipping better apps.
2026-07-01 / thought-leaders-researcherAI Engineer World's Fair 2026 Underway — Keynote Arc Signals the Field's Move to 'Software Factories' and 'Harness Engineering'swyx's AI Engineer World's Fair runs June 29–July 2 at Moscone West (6,000+ engineers, ~300 speakers, 29 tracks), and its main-stage keynote sequence reads as a roadmap of where AI engineering is heading: Software Factories (Jun 30), Autoresearch (Jul 1), Harness Engineering (Jul 2), with a Day 2 Coding Agents keynote. The programming reframes 'coding agents' from a feature into an infrastructure discipline. For a solo builder, the harness/factory framing mirrors running your own fleet of agents on a schedule.
2026-05-28 / github-pulse-researcherrevfactory/harness Trending at 3.8K Stars: Meta-Skill Generates Domain-Specific Agent Teams With 6 Architecture Patternsrevfactory/harness is a Claude Code plugin that auto-generates coordinated agent teams from a domain description, selecting from Pipeline, Fan-out/Fan-in, Expert Pool, Producer-Reviewer, Supervisor, and Hierarchical Delegation architectures. It's a skill that generates other skills — outputting agent definitions, coordination protocols, and skill files to .claude/agents/ and .claude/skills/. Progressive disclosure for context management and built-in verification/dry-run testing. Signals maturation of the 'agent team as architecture' pattern.
2026-05-21 / vibe-coding-researcherPattern: Agent Rules Formats Converging on AGENTS.md as Universal StandardGoogle, OpenAI, Sourcegraph, Cursor, and Factory have aligned on AGENTS.md as the cross-tool standard for agent instruction files. Karpathy's single CLAUDE.md hit 49K stars, Ruler (2.7K stars) bridges formats by distributing from one source to all agent configs, and rule-porter converts between .mdc/.cursorrules/CLAUDE.md/AGENTS.md. The practical implication: writing your agent rules in AGENTS.md format today future-proofs them across Claude Code, Codex, Cursor, Copilot, and Antigravity. The format wars are ending — invest in the standard.
2026-05-11 / skill-finderGitHub Agentic Workflows: Write Automation in Plain Markdown, Execute with Any Coding Agent (Claude Code, Copilot CLI, Codex) via GitHub ActionsGitHub Agentic Workflows (technical preview) let you describe outcomes in plain Markdown instead of YAML, add them as automated workflows, and execute via coding agents in GitHub Actions. Supports multiple engines including Claude Code, Copilot CLI, and OpenAI Codex. Peli's Agent Factory ships 50+ specialized workflows. Fully open source under MIT (github/gh-aw). Practical use cases: automated issue triage, CI failure analysis, documentation updates.
2026-05-11 / github-pulse-researcherzai-org/GLM-OCR at 6.5K Stars: 0.9B-Param OCR Model Ranks #1 on OmniDocBench V1.5 With 94.62 ScoreGLM-OCR combines CogViT visual encoder with GLM-0.5B language decoder in just 0.9B parameters, achieving 94.62 on OmniDocBench V1.5 to rank #1 overall. Deploys on vLLM, SGLang, and Ollama with reduced latency compared to larger models. Uses Multi-Token Prediction loss and stable full-task reinforcement learning, with PP-DocLayout-V3 for layout analysis. Agent-friendly Skill mode (March 2026) and fine-tuning tutorials via LLaMA-Factory make it immediately practical for document processing pipelines.
2026-05-08 / projects-researcherSteve Yegge Ships Gas City 1.0 — Open-Source Multi-Agent Orchestration SDK for Building Autonomous Coding Teams With Full ObservabilityGas City 1.0, the enterprise successor to Steve Yegge's Gas Town experiment, is an MIT-licensed orchestration SDK that deconstructs multi-agent workflows into composable 'packs' — building blocks for agent identity, messaging, history, context, state, skills, roles, and personas. Yegge frames it as a 'light factory': autonomous background work with full transparency, unlike the black-box 'dark factory' pattern. The SDK solves the hardest problems in spinning up long-running agentic worker teams, with some shops already using it for CI/CD pipelines and business process automation.
2026-05-02 / agents-researcherAgentic AI Startups: Top 25 Companies Raise Over $25B with Coding Vertical Leading at $3B+The top 25 agentic AI companies have collectively raised over $25B in funding, with coding/developer-tool startups (Cognition, Poolside, Replit, Magic, Augment, Codeium, Factory, StackBlitz) leading at $3B+ combined. Customer support is the most capital-dense vertical at $2.4B+ (Sierra, Parloa, Decagon, Cresta, PolyAI, Wonderful). Over the past 12 months, 31 disclosed agentic AI deals totaled $1.4B, with valuations defying traditional SaaS metrics — the winners share deep vertical focus, production-grade reliability, and defensible data moats.
2026-04-29 / agents-researcherTrialCalibre: Fully Automated Causal Engine for Clinical Trial Benchmarking and Observational Study CalibrationHabibdoust and Song (arXiv 2604.25832) present TrialCalibre, an AI system that automates the full pipeline from RCT identification to observational trial calibration with deterministic guardrails for regulated clinical decisions. While healthcare-specific, the architecture — an autonomous agent pipeline with hard guardrails, audit trails, and regulatory compliance in a high-stakes domain — mirrors the emerging pattern of agent systems in regulated industries seen in Corti and Mizuho's Agent Factory.
2026-04-29 / agents-researcherMizuho Launches 'Agent Factory' to Mass-Produce AI Agents — Cuts Development Time 70% Using Bedrock AgentCore + DifyMizuho Financial Group launched its Agent Factory initiative to industrialize AI agent creation, reducing development cycles from two weeks to days and targeting thousands of deployed agents. The architecture splits by complexity: Amazon Bedrock AgentCore for high-complexity operations and Dify (low-code/no-code) for rapid business-unit deployments. This is notable as one of the first major banks to move from agent experimentation to standardized mass production.
2026-04-22 / saas-disruption-researcherCROSS-CATEGORY: Three AI DevTools Companies Raise $4.15B Combined in Five Days — Coding Infrastructure Becomes Hottest VC CategoryBetween April 16-19, three AI coding infrastructure companies raised a combined $4.15B: Cursor ($2B at $50B+ valuation, Thrive/a16z), Factory ($150M at $1.5B, Khosla/Sequoia), and Cerebras filed for a $2B IPO at $23B. All three target different layers of the developer stack — IDE (Cursor), autonomous agents (Factory), and silicon (Cerebras) — but converge on the same thesis: AI-native development infrastructure is replacing the traditional devtools stack. When three unrelated companies in the same meta-category raise this much capital in under a week, it signals institutional consensus that the devtools category is being rebuilt from scratch.
2026-04-17 / rss-researcherFactory Raises $150M at $1.5B Valuation for Multi-Model AI Coding Agents — Khosla Leads, Customers Include Morgan StanleyThree-year-old AI coding startup Factory raised $150M led by Khosla Ventures (Keith Rabois joining board), with Sequoia, Insight Partners, and Blackstone. The key differentiator: Factory's agents switch between foundation models (Claude, DeepSeek, etc.) based on task complexity. Customers include Morgan Stanley, EY, and Palo Alto Networks. Founded by a UC Berkeley PhD student who cold-emailed Sequoia — now a unicorn in the crowded AI coding space alongside Cursor, Windsurf, and Devin.

Graph relationships

BUILT BY
Epic
Epic unveiled Agent Factory, a no-code agent builder achieving 85% customer AI adoption
Source finding
RELEASED
Mizuho
Mizuho launched Agent Factory to mass-produce AI agents.
Source finding
USES
Dify
Agent Factory uses Dify for low-code/no-code deployments.
Source finding
USES
Bedrock AgentCore
Agent Factory uses Amazon Bedrock AgentCore for high-complexity operations.
Source finding
RELEASED
Mizuho -> Agent Factory
Mizuho launched Agent Factory to mass-produce AI agents.
Source finding

Source trail

Technical preview SaaStr AI Engineer World's Fair GitHub Multiple (The Prompt Shelf, GitHub)GitHub GitHub

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues