Daily Brief

Full episode: https://www.youtube.com/watch?v=OS1NZLgKM2c Me on twitter: https://x.com/dwarkesh_sp

Decagon builds the customer support agent that answers when you email a brand, and making that agent good for a specific enterprise is the forward deployed job. It splits in two: one side configures the agent brain, the instructions and the handoff rules for when a human should take over; the other works like product engineering, figuring out what a new enterprise needs and anticipating the requests that have not been made yet. Sunny Rekhi's point is that the line between forward deployed and product is thin, because a customer ask is often just a product feature waiting to be built. What cha

At Palantir, forward deployed started as a literal description: you were deployed, physically, at the customer's site, and the onboarding project was keeping the platform from falling over. Natalie Meurer's dirty secret is that the title never settled after that. It stretched across DevOps, data integration, ontology work in Slate and then Foundry, and solution architecture, until forward deployed engineering meant so many different jobs that the label stopped meaning much. Every company hiring for it, whether they call it forward deployed, customer, or deployed engineering, is describing a sl

It's Friday night, an enterprise sales rep needs an SAP S4 HANA integration to hit quota, and the reflexive Forward Deployed answer is yes. Leo Mehr's first principle is to pause instead: always be scoping. Saying yes to everything buries the team and often does not even serve the customer, so the job is to weigh what actually matters against the rest of the queue and decide with that context. Ramp's FDE function looks different from its Palantir origins, pointed at enterprise customers, but the discipline is the same: scope hard before you roll up your sleeves and ship. The second half is wh

Most of the coding agent market quietly optimizes for token usage; Cognition's deployed engineering team measures the opposite, the outcomes a customer can actually see, and reports something like an 82% reduction on the work they targeted. Jia Wu's argument is that you measure before Devin ever lands, then again once it is fully activated inside the customer, so the value is a real delta and not a vanity number. The way that value shows up is not linear: one team using the agent is a step function, a whole enterprise using it is parabolic, because the products Cognition builds and the custome

Full episode: https://www.youtube.com/watch?v=OS1NZLgKM2c Me on twitter: https://x.com/dwarkesh_sp

Jerry Tworek led reasoning at OpenAI, convinced that scaling reinforcement learning was the path to AGI. Rohan Anil co-led Gemini pre-training and built the Shampoo optimizer. Now they've teamed up at Core Automation on a contrarian premise: the transformer has carried us as far as it can, and the bottleneck to smarter systems is no longer scale — it's the architecture itself. The missing capability is continual learning, models that adapt at test time, which transformers can't do. In-context learning taps out fast (Codex needs compacting after ~20 minutes) and fine-tuning invites catastrophic

Since skills were open sourced, Yogendra Miraje's team at FactSet stopped thinking about shipping features and started thinking about shipping skills. A skill is a capability you hand the agent, and its heart is a short skill.md whose name and description are really routing signals: get them distinct and the agent triggers the right one, blur them and it fires the wrong skill or none at all. He walks through a minimal skill registry, progressive disclosure so the agent only loads what it needs, and trigger words, like asking for a PDF versus an HTML report, that decide which skill runs. The h

Nubank serves 135 million customers, so an AI agent that mishandles a support conversation fails at scale. The talk opens with the result: five agents in production, higher customer satisfaction, and roughly 20 times faster shipping. Shreya Rajpal, CEO of Snowglobe, argues the thing that unlocked that pace was evals, and specifically simulated data standing in for real conversations. Good agent evals are hard because the data is multi turn and stateful, not single turn question and answer, and hand curating it and waiting on production to confirm can take forever. Snowglobe points at the agen

Ask a frontier model how to make a rental property more profitable and it will tell you, with total confidence, to raise a good tenant's rent by 5% to 10%. Udi Menkes calls this the fluent bluff: advice that reads perfectly and is quietly wrong, because the model has fluency but no experience. He points to a Princeton study where models handed a million dollars to run a business barely turned a profit, and to his own tests where off the shelf advice would have cost real customers real money. His argument is that context is not experience, and the fix is grounding. At Intuit he turns real busi

A team ran about a thousand people through a market research survey, then had LLM agents replay the same questions, and the agents matched the humans closely while carrying less noise than the humans did themselves. That is exactly the tell: a synthetic respondent smooths over the messiness that makes a real population real. Ishan Anand walks through where it breaks. Nudge one variable in the prompt template and purchase probability swings, because the model infers latent confounders nobody stated, a little like it is playing improv. So how you ask matters as much as which model you pick, and

From building no-code products at Airtable to leading Core Product Engineering at OpenAI, Akshay Nathan has spent much of his career trying to make the power of software accessible to people who do not write code. In this episode, Akshay joins swyx and Vibhu to unpack the launch of ChatGPT Work, why Codex unexpectedly took off among nondevelopers inside OpenAI, and the company’s broader plan to bring useful agents from software engineers to knowledge workers and eventually everyone. We go deep on the shared agent harness behind Codex and ChatGPT Work, why OpenAI brought the experiences togeth