AI hacking capabilities that were restricted by frontier model guardrails just months ago are now available through open-weight models.
Factory's @varinnair warns attackers are already using custom AI agents in real-world cyberattacks:
"The open weight models have clearly caught up in terms of what they can do in terms of offensive security. Now these tools are available to anybody."
"People have started building custom harnesses like ARTEX and HexStrike, which are used in real attacks today, in the last few months."
"Mythos really scared a lot of CISOs because they realized that, okay, theoretically, models can find vulnerabilities in your system, but they hadn’t really seen a breach. That changed in October, on Black October."
"After Black October, people suddenly started realizing, 'Okay, this is actually a big deal. We need to do something with our budgets.'"
@factory
Factory is a product company. In a year when a lot of smart money is betting on neoservices that is starting to sound contrarian.
The AI wave makes it tempting to sell outcomes and deliver them with people. We think the bigger opportunity is software that gets better every time it's used. That's what we build for our customers, and it's how we run Factory.
The products we build should help our customers build their own capacity to ship better software, faster. We think of the product as everything from the first time you hear about Factory through your Nth renewal.
Building the right product starts with the old advice: listen to your customers' problems, not their solutions. In 2023, many people believed we should have built autocomplete. We understood that we instead had to build toward self-improving software and software factories, since that is what would come next.
That is our product goal: self-improving software, which continuously improves from the data feedback loop of its users. Self-improving software doesn't happen on its own. You have to be data-backed and optimize the signal → deploy loop. That requires a software factory operating continuously and evolving alongside the software it produces.
The factory also has to know what good looks like. It has to help the organization extract its tacit knowledge and embed it into the software that operates the business. Taste is not sufficient. You must enforce the unique decisions that define your product. And it doesn't stop at the product. The entire company is going to be defined by software that listens to the rhythm of the business and enforces it. The clear goal of a leader is to ensure the quality and correctness of these systems.
We run Factory this way. Our unique product and technical decisions are codified as customizations to our harness, automations, CI/CD, lint rules, etc. Factory agents sit in our shared Slack channels, where they build context and execute tactical work on demand.
It also changes how we organize. Your Engineering, Product, and Design team will ship your org chart, so you must design an organization where incentives are maximally aligned. Our pod structure maps directly to our product taxonomy, which gives each pod deep expertise and agency. We work carefully to make sure the product feels cohesive and not like multiple disconnected teams. This is not easy, and we still haven't perfected it. We keep the minimal number of layers between leaders and the organizational lines. 1:25 is not unreasonable for a true people manager.
Our product operating system is itself self-improving software. It aggregates feedback, tracks launches across the team, and organizes our work. With it, our Head of Product @TaylorTheSavage runs a product surface that would otherwise require an army of PMs with just a handful of high-ownership, high-leverage ones.
When an org is designed this way, its impact scales with software, not headcount. That is the real difference between a product company and a services company. Selling outcomes is right. Delivering them with people is services. Neoservices arbitrage the gap between how fast the technology moves and how fast buyers adopt it, and some will become large businesses. But that gap closes every time a product company sells the same buyer a better long-term solution. The market already knows how to value services: Infosys does about $20B in revenue and trades at roughly 2x, not 50x, and it's down about 40% this year.
Being a product company is harder. @danlovesproofs once told me something I very much agree with. Product companies are like movie studios. You have to keep producing blockbusters, and that helps you build talent, brand, and resources. Gone are the days when you could coast on one hit. Do that, and you'll be acquired by Bending Spoons at a fraction of your value. The war is always being waged, and no winner is ever truly ahead by much. And demand for the movies is up.
That's true even against the model labs. There is a value gap between the models and outcomes for users. Anyone can close that gap, and there's no reason to believe any company has an automatic right to win the products that matter most. If anything, access to the full pareto frontier of cost and quality gives model-independent, product-focused companies a focus, cost, and quality advantage.
If someone asked me what to build today, I'd say go build a product company. It is harder, but the biggest outcomes are only possible through product.
Custom Automations are now GA in Factory.
Describe a recurring workflow, pick a schedule or event trigger, and Droid runs it to the intended result. You choose the model, the machine, and Connectors for each automation.
We are terminating Chris Degnan for unethical conduct involving Cognition.
The last few months have seen incredible progress in AI capabilities. San Francisco has flourished as new companies that solve new, more ambitious problems are finding great success. Generally, it is a wonderful time to be building.
We at @FactoryAI have seen overwhelming interest in our model-agnostic coding agents and our software factory product. Our team has 10x’d in size, while our revenue has 100x’d year over year. This momentum has been unprecedented.
A much larger competitor, Cognition (makers of Devin) has fallen behind us on the capabilities that matter most to customers: cost, quality, and security. Instead of competing in the market, Cognition engineers feigned interviews with us to pry information about our product. Not finding what they were looking for, Cognition has decided to throw their weight and money at people with direct knowledge of our most confidential plans.
It feels as though ethics is being thrown out of the window in the AI era. People are willing to do anything, including exploiting privileged information and violating ethical boundaries. I think this is unique to our time, and I don’t think it’s right. Integrity still matters.
Yesterday, I made the decision to immediately terminate Christopher Degnan’s roles as a Board Observer and Advisor to Factory, after over a year of service.
Prior to this decision, Chris told me he had a casual conversation with an executive at Cognition AI. When I questioned his intentions, he reassured me that ethics aside, he had “made too much money” and was “too lazy to go work for Cognition,” which I trusted and believed.
On Monday, Chris spent time advising the Factory team on a handful of confidential board-level matters. That evening, he called me to say that the conversation that was initially described as casual and one-off was actually formal and recurring. For weeks, while he sat in our board meetings and advised our leadership team, he was also confiding with executives of our largest competitor. Chris was subject to confidentiality obligations in connection with his work with Factory. We do not know the extent of the information he shared, but it puts his timely questions about our product roadmap and what the parity gap involves into a new light.
It is sad to sever a relationship with someone who has been a trusted advisor - and even a close friend - for over a year. But Chris’s conduct is unacceptable to me. Trust in Board Membership is one of the sacred bonds in the Silicon Valley, one that helps the startup ecosystem thrive. With it comes an enormous responsibility. That trust was violated.
Competition is good. I respect and in many cases admire our competitors. San Francisco is a beautiful, singular place where the bold and ambitious go to defy the norms and precedents of the past. But certain principles that must remain. Violating ethics to seek advantage turns what should be positive-sum into zero-sum.
We all love technology. And we all love to compete. But we must hold ourselves to a higher standard. The future of software engineering comes with abundance that will impact every person on Earth. Building that future comes with immense responsibility. We must build and compete with integrity.
Many are surprised to hear that Factory is the first and only on-prem Software AI product that exists today. One with the ability for the control plane to be hosted in the customers' VPC or on-prem environment.
These are the real challenges to AI adoption in the enterprise. We are proud to support a best-in-class experience for customers' to own their intelligence while deploying AI at scale.
We have raised $200M at a $5B valuation to scale self-improving software development in the enterprise.
@FactoryAI has grown to serve hundreds of thousands of developers at companies including RBC, Adobe, Nvidia, T-Mobile, and Palo Alto Networks.
We will use this capital to accelerate our investments in research, product, and global go-to-market.
10,000 human-days of modernization work done in 45 days.
45 day database migration completed in 6 hours.
See how @Comarch is building its own software factory.
Most startups celebrate their first couple million of revenue. @FactoryAI gave it back.
They didn’t have to. They chose to. They decided that the product just wasn’t good enough yet, and they wanted to know that they’d really earned it. They wanted their customers to be obsessed.
Two years later they shipped Droid CLI, and they’ve been ripping ever since. The product is fantastic and the team is even better - particularly having been hardened by their “two years in the desert”.
@matanSF was one of our very first guests on Training Data, and we are delighted to welcome him back.
00:00 Introduction
01:37 Enterprise Lock In Fears
04:26 No Lock In Promise
05:24 Two Years Early
07:24 Refunding Revenue
13:49 Droid CLI Breakthrough
16:44 Harness Frontier Tactics
26:42 Natural Language Routing
27:32 Open Models Catch Up
29:26 Token Share Forecasts
31:02 Automating Low Leverage Work
32:36 Pricing Beyond Tokens
35:56 Software Factory Vision
43:43 AI Transformation Playbook
46:13 Async Agents And Optimism
We’re seeing model routing deliver significant savings across large enterprises and all workflows. Full model independence and reliable harness-level routing enables this access to the Pareto frontier.
Excited to share this work with everyone! More to come. Expect more savings as efficient models and routing continue to improve
This is exactly why we believe in customization.
Quick context: Factory's original secret scanner was deterministic, so it either flagged things that weren't actually secrets (false positives) or missed secrets that didn't match its patterns (false negatives). Their solution was elegant: two small post-trained models—one to catch missed secrets and another to filter out false alarms.
The result: models that outperform GPT-5.5 and Opus 4.8 on this specific task while running at a fraction of the cost and latency.
This is our thesis in action: take a strong open model, post-train it for a specific production problem, and you can build something that's faster, cheaper, and better than frontier models at that job.
Congrats to @FactoryAI. Proud they built Droid Shield 2.0 on @FireworksAI_HQ.