Factory

Opus 4.5, meet Droid. Available to all Factory users at Sonnet prices for the next two weeks. The new default model in Factory, Opus 4.5 pushes the frontier of general intelligence and is our new daily driver for software development. The model shines on larger or more ambiguous tasks that require deeper research and thoughtful planning, including migrations, refactors, and large-scale feature or prototype development. Opus 4.5 introduces a new effort parameter, which controls how much thinking the model does per task. As a highly capable model generally, this allows significant control to trade off more intelligence for lower cost and latency. We recommend using High to get the most out of this model, particularly when speccing, debugging, and iterating on complex tasks interactively.

We are excited to have Andy Taylor join Factory’s GTM team. Andy joins us from Anysphere / Cursor. Andy has spent his career selling to developers and deeply technical engineering teams, building early sales motions at companies like Teleport, Temporal, and Baseten. Andy joins our killer GTM team and will be working with distinguished enterprises to bring agent-native development and the Droids into production. Welcome!

Drew and I recently made a jingle for one of our portfolio companies, Factory, for their droids… which sounds simple enough, but it’s kind of hilarious how much thought we put into something that lasts literally half a second. I won’t lie, I’ve always been envious of the people who got to create the Netflix “badoom” or the McDonald’s jingle. I know Finneas is doing Apple TV’s right now, which he will no doubt nail… But anyway, back to us! So we called it “Beep Boop,” which feels about right. Making a jingle is a totally different process from a full song, which yes, sounds painfully obvious. You can’t think about melody arcs or structure, you’re just trying to make something that feels unique, gives the product some identity, and doesn’t annoy the hell out of everyone who hears it a hundred times a day. You have to consider things like is this a sound that defines success? Completion? Warning? Failure? Money? Thinking? Etc. etc. etc…. And there’s a fine line there, because you want it to stand out, but not in a “please make it stop” kind of way. We actually went down a deep rabbit hole of all Apple’s sounds, and they really have created an orchestra of amazing sounds that all acutely feel like the brand. All that said, I’m proud of what we came up with and glad to see it out in the world. Didn’t think I’d ever say we produced a track for robots, but here we are. Add this to our repertoire: Factory droid completion boop and incident.io alert sound. Maybe there should be a Grammy category for this after all… If you use Factory, check out Completion Sound FX-ACK01!

We detected and disrupted a highly automated cyber operation attempting to use Factory as a node in a worldwide mesh of “off-label” LLM usage. The attackers deployed AI coding agents to generate and maintain their infrastructure, adapt to our defenses in real time, and orchestrate traffic from tens of thousands of synthetic organizations. This attack mirrored similar incidents across the industry, including those recently disclosed by Anthropic. The level of automation and speed we observed have direct implications for how state‐linked and organized cybercriminals will operate moving forward. We're sharing our response so others can strengthen their defenses. Read more at the link in the comments.

We’ve partnered with Palo Alto Networks to bring Prisma AIRS directly into Factory’s agent-native development platform. Now every prompt, model response, and tool call can be inspected for prompt injection risks. “This partnership brings two categories together: agentic development and autonomous security. By integrating Factory’s Droids with Palo Alto Networks Prisma AIRS, we will give organizations the power to build with AI while staying fully protected in real-time.” - Matan Grinberg If you are interested in securely deploying agents to your enterprise, reach out to learn more about Droid Shield powered by Prisma AIRS. 

Gemini 3.0, meet Droid. We’ve been hands-on with the Google DeepMind Gemini team over the past few weeks, and Gemini 3.0 has impressed us with: - End-to-end tasks with continued effectiveness on long contexts - Proactive execution with faithful state tracking and adherence to plan - Strong balance between interactivity and agency with dynamic reasoning We're offering Gemini usage at a discount: our 10M free trial now grants you 20M Gemini credits to try the new state-of-the-art model inside the #1 software development agent. We're excited to deepen our partnership with the Google Cloud team. More on this soon.

We just launched the largest usage plan in all of AI. Last week, we announced the Factory Ultra Plan. $2,000 / mo for 2B Factory Standard Tokens. applicable across Sonnet 4.5, GPT-5.1, Opus 4.1, and more. Lots of debate about whether AI spend is “real” or just another VC-subsidized bubble. Here’s what we’re actually seeing at Factory: Lightweight, incremental AI tools (which are priced accordingly) have flooded the market. But our customers have entered a new phase, where AI agents (our Droids) are literally driving organizational augmentation. These small teams want systems that take on entire workflows and meaningfully change operational throughput. Small teams that don't have time to wait for hiring pipeline and onboarding hit the limits of our Max plans within weeks, and kept asking for more. Users are willing to pay more because AI is finally creating value at the scale of actual work. The productivity our customers unlock isn’t measured in minutes saved or lines of code generated. It’s measured in real business outcomes. Revenue, speed, and headcount leverage.

GPT-5.1 meet Droid. We've been working with OpenAI testing the new GPT-5.1 and GPT-5.1-Codex models over the past few weeks, and here’s you need to know: • GPT-5.1 is designed for active collaboration. Great for brainstorming, planning, writing, quick architectural decisions, and pair-programming when you want a model actively in the loop. • GPT-5.1-Codex excels at long-running agentic coding tasks, broad edits, migrations, and PR level changes. Both models perform well on adaptive reasoning and code quality. Both models available now in Factory.