Factory AI

Factory mined 780,000 web searches its Droid agent ran in production to see what coding agents seek outside the codebase, from API docs onward — with the agent itself writing the queries, pipeline, and charts.

Factory's AutoWiki generates and refreshes a structured, source-grounded wiki for a repository through a multi-agent pipeline (survey, plan, generate, capture, publish), giving engineers and agents a shared, current map of a codebase's architecture and flows.

Factory open-sourced Droid Shield 2.0's secret-detection models, two fine-tuned Qwen 3.6 35B A3B models for catching missed secrets and cutting false positives, releasing LoRA weights on Hugging Face, and used the same review pipeline to disclose CVE-2026-42876.

Factory lays out why coding agents underperform on real codebases: models get task descriptions and files but miss the historical, architectural, and collaborative context human developers rely on, and describes the scaffolding layers it built to close that gap.

Factory launched revamped Analytics for its coding agent, breaking down consumption, model efficiency and adoption by model and user across all organization sessions, aimed at cost visibility for agentic engineering teams.

Factory details how it detected and dismantled a large, likely China-linked fraud operation that used AI coding agents to generate and adapt infrastructure in real time, chaining free-tier LLM access across tens of thousands of synthetic organizations to resell and mask further abuse.

Factory shipped redesigned Analytics showing credit consumption by model and user, cost concentration, adoption metrics and estimated router savings, with an Analytics API for BI export on Business and Enterprise plans.

Factory's Missions breaks large multi-day projects into milestones and features, spawning fresh worker sessions per feature with clean context, coordinating handoffs through git, and validating at every milestone instead of running one long agent session.

Factory's Droid benchmarks show its model-agnostic agent scaffolding beat labs' own CLIs on Terminal-Bench across every model tested, arguing that agent design (prompting, tool design, environment exploration) drove more performance gain than model selection.

Factory makes Automations generally available: describe a recurring workflow, pick a trigger, model and machine, and Droid runs it. Templates cover ticket-to-PR, code review fixing, security audits, CI triage and Slack briefs.

Factory.ai's research post details an upgrade to its autonomous-coding secret-detection system, adding two fine-tuned models that sit on either side of a deterministic pattern scanner: one 'risk' model catches secrets the scanner misses, and a 'downgrade' model clears false-positive flags by reasoning over a scrubbed code window. Trained and evaluated on Samsung's CredData benchmark, the models reportedly match or beat frontier models on both tasks at much lower cost and latency, and Factory is releasing the model weights publicly.

Factory's Incident Response connects an agent to a Slack alerts channel from tools like Sentry, Rootly, or Datadog; when an alert fires, the agent investigates, triages, and drafts a fix, building a persistent runbook that improves future responses.

Factory made Delegations generally available. Admins configure the default model, instructions, plugins, service account and compute once, then team members delegate tasks to Droid on remote machines, running under separate Git credentials and connectors. The video was not transcribed.

Factory joined an OpenAI-convened working group (with Google, Sourcegraph/Amp) standardizing AGENTS.md, a root-level file giving any coding agent build, test, style, and security instructions instead of vendor-specific config files.

Factory announced its coding agent is now available in Atlassian Jira, letting teams assign a task to it as they would to a teammate. Based on the announcement post only; install steps, permissions and agent behavior were not covered in the source.

Factory's /readiness-report scores a repo across eight pillars (linting, build system, tests, docs, dev environment, code quality, observability, security) and five maturity levels, showing concretely why agents struggle on under-instrumented repos.

Factory's account of moving past descriptive agent analytics: capturing failure signals from live sessions and closing the loop so the agent implements its own fixes, with the argument for why conventional metrics can't drive that.

Factory reports its model router now cuts production inference costs by 63% versus frontier pricing, up from 42% in June, while matching 99% of Claude Opus 4.7's Terminal-Bench 2 pass rate, and is expanding to an on-premises version.

This Factory.ai blog post argues that as AI coding agents take over more of the actual code writing, linters should be used to encode architectural rules, boundaries, and style conventions as machine-verifiable constraints rather than relying on human code review or tribal knowledge. It frames lint rules as a more reliable control mechanism for steering agent-generated code than traditional guardrails.

Factory introduced OpenCompute, a private-preview orchestrator that provisions and manages agent sandbox fleets on customer-controlled infrastructure, building on Remote Control and tying into Automations and Environments.

Every multi-step agent run eventually outgrows its context window. Factory walks through the strategies for deciding which prior turns to keep, which to summarize, and which to drop, and how those choices show up in quality, latency, and spend.

A Factory.ai incident report describing how the company detected and shut down an automated attack campaign that tried to hijack its AI software development platform, using it as a node to resell or launder off-label LLM API access at scale.

Factory added support for running its agent platform against a local NVIDIA DGX Spark deployment of Nemotron 3.5 Lightning (a 30B MoE model, 3B active parameters), letting security-sensitive teams keep source code and execution inside their own environment.