Out of the box, long-horizon agents struggle to accurately perform end to end work in the real economy (outside of coding) because those tasks are not easily verifiable, the data is hard to scale, and going from inputs to real outcomes can actually take many days. Even if you had a reliable way to verify outcomes at scale (and weren’t bothered by the multi-hour iteration loops), the sheer volume of decisions by the agent that occur in a multi-hour job makes it hard to know whether performing well will generalize to production. Over the last two years at @trybasis, we've been solving this problem by supervising the process our agents take to get to outcomes, rather than just looking at whether the outcome itself is correct. We think this is the key to building production agents at scale. It's what has allowed us to run agents in production that operate for hours, sometimes days, and reliably perform tasks like entire complex tax returns end to end. Today, alongside @braintrust, we're open sourcing a standard for defining, evaluating, and eventually rewarding agent behaviors. Thread below with all the details on how we’re scaling behaviors to close the loop for long-horizon agents.

Excited to announce this milestone. We are deploying long-horizon agents at scale that can do real work in the real economy. Our agents can already operate reliably for hours at a time and I expect that to be days by the end of the year. We can barely keep up with demand. If you want to come join and work at the frontier for the most intense, fast-paced couple years of your life, send me a note.

It's been great working with @OpenAI to build frontier accounting agents that are increasingly taking on more complicated and longer running workflows. If you're looking to work at a company deploying agents in production to automate material parts of GDP, shoot me a dm

If you used to work at an NYC tech company and want free 10k: Just DM me a note with top 5 people you've worked with in eng/product. No intro needed. just 5 names and if we hire any of them in the next 12 months I will pay you 10k. You don't need to do any work.

Introducing Deployed Intelligence Intelligence is becoming abundant. Putting it to work remains hard. A new era demands a new approach For the first time in history, we can give every accountant the equivalent of a 50-person team to support their work. A team of capable professionals that is instantly available and always-on. But this sudden leap in capability comes with a challenge: we’ve changed the tools to do the job, but not the mental models. We’re asking accountants to go from doing the work to directing a team of agents… and we need to teach them how! This is the silent failure mode of most enterprise AI efforts: assuming that adding AI to existing workflows will just work. It won’t. Empowering users with AI demands new ways of thinking about work. The AI revolution will be deployed, not just built Everyone is watching the progress frontier models are making with great interest and rightly so. Fewer people are focused on the messier, perhaps more important challenge: getting the intelligence into the hands of users to do real work. We continue to believe that AGI will arrive faster than people think (short timelines) but will take far longer to diffuse (slow takeoff). The friction is not only technical; it’s human and organizational. It’s legacy software. It’s institutional habits. It's rigid workflows designed in an era where intelligence was scarce. And that’s why deployment matters. Over the last two decades, we’ve seen several waves of software deployment models, from on-prem to SaaS to bespoke platforms. With each wave came new deployment roles: implementation managers, solution engineers, and forward-deployed engineers. Now, we believe a new phase is emerging: one defined not by implementing software, but by deploying intelligence. Deploying intelligence will be a defining challenge In an age of abundant intelligence, simply having AI isn’t enough. The advantage goes to those who learn to wield it well. This applies at every level: Individuals who learn to collaborate with AI will dramatically outperform those who don’t Companies that embed intelligence into workflows will outpace those who merely bolt it on Solution providers who master deployment and make intelligence usable will lead the way Deploying intelligence is challenging because it is about fundamentally changing the way work gets done, rethinking how users interact with software, and embedding AI so deeply into workflows that it feels like second nature. That’s why we’ve formed the Deployed Intelligence Team at Basis. What does deploying intelligence look like in the real world Deploying intelligence means effectively guiding organizations through the shift from software-driven work to intelligence-driven work. The Deployed Intelligence team is the group on the front lines; we parachute into our customer’s organizations, learn their unique context, and give them the tools to redesign their accounting workflows around intelligent agents. It means creating a future where AI isn’t an add-on, but rather the backbone of how accounting work is done. So who will do this kind of work? It’ll be someone who looks at how things are done today and says, this doesn’t make sense, it can be done better. Someone who learns new technology faster than anyone else on their team, and then feels compelled to teach it to everyone around them. If that’s you, you’ll feel right at home here. Whether you’re an accountant or not, this team is for people who understand the future of knowledge work and want to be the ones creating it. Shape the Future with Us If you’re excited about deploying intelligence into the real economy then come join us