
Summary
Written by a model from the transcript; not yet reviewed.
In this episode of the a16z Podcast, hosts Marc Andreessen and Ben Horowitz sit down with TypeSafe AI founder Diogo Almeida to discuss why current AI coding tools are not making software inherently better and how a new primitive called Jev aims to change that. Diogo argues that while coding agents like Claude Code and Codex make it faster to produce traditional software, they merely automate 10-line pull requests without expanding what software itself can actually do. Instead of replacing software engineers with faster, sometimes worse code generation, TypeSafe wants to empower developers by embedding probabilistic intelligence and state machines directly into programs. The discussion explores Diogo's unorthodox journey from a mathlete and Kaggle competitor to OpenAI, the philosophy of building reliable production software over sci-fi hype, and why the future of AI in computing looks more like an inverse SaaS apocalypse than a software-pocalypse.
- Coding agents speed up writing code, but they do not make software itself more capable. 0:07
- Jev functions as a state machine combined with natural language, giving programmers a new primitive to expand software capability. 4:04
- The guest prioritizes intelligence per dollar and relentless reliability over chasing abstract AGI milestones. 8:22
- Building reliable background automation requires statistical guarantees so systems can run without constantly paging developers. 16:15
Chapters
- 0:00Podcast Intro and Automation Frustration
Hosts introduce the episode and lament how AI, despite its intelligence, fails to automate many software tasks.
- 0:48The SaaS Apocalypse and Jev’s Upswing
The hosts discuss how Jev could reverse the perceived decline of SaaS by adding true intelligence to existing software.
- 1:23Reliability Over Showmanship
The conversation shifts to why reliable, deterministic behavior is more valuable than flashy demos for production AI.
- 6:51Personal Journey into AI and TypeSafe
Diogo recounts his path from mathlete to AI researcher, highlighting the moments that led him to build TypeSafe.
- 12:01Future Applications and Probabilistic Programming
The panel speculates on Jev’s potential to enable new AI‑driven features and to usher in a new era of probabilistic programming.
- 26:37Defining Reliability for AI‑Powered Software
Diogo explains the multi‑dimensional concept of reliability—uptime, determinism, and robustness—in the context of Jev.
- 33:04Closing Reflections and Call to Action
Diogo concludes with a vision of software that “does what I mean” and invites listeners to engage with the episode.
Highlights
- 0:07
Software Isn't Getting Betterinsight
The guest points out that despite the widespread use of AI coding agents, software is not actually getting better and might even be getting worse.
- 4:04
Smart Software Versus In-Time Softwareinsight
The guest explains that coding tools provide in-time software on the fly, whereas Jev is designed to expand what software itself can do.
- 11:09
The guest shares how winning a Kaggle competition through brute-force systems automation rather than sophisticated math led to being adopted by SVM co-inventor Isabelle Guion and speaking at NeurIPS.
- 18:22
Testing Generalization at OpenAIstory
The guest describes testing early RLHF models at OpenAI with unusual queries like eating socks before meditating to prove the model was truly generalizing rather than just cheating.
- 27:06
Defining Reliability as Robustnessstrategy
The guest explains that reliability in Jev means robustness, ensuring the model delivers consistent intelligence even if the exact output is not strictly deterministic.
- 31:42
The Inverse SaaS Apocalypseinsight
The guest argues that rather than destroying SaaS companies, AI will make existing SaaS platforms dramatically more useful by automating complex workflows.
- 40:10
The Five Stages of AI Griefinsight
The host describes the five stages of grief developers go through when trying to embed traditional LLMs into software before finally giving up and routing output to humans or while-loops.
Quotes
“It doesn't matter how much AI coding agents you use, the software actually isn't getting better.”
Martin Casado · 0:07
Sponsored segments
- A16Z 42:13–43:02
About this episode
a16z’s Ben Horowitz and Martin Casado sit down with TypeSafe AI founder Diogo Almeida to ask a simple question: AI has become remarkably capable, so where is all the automation? Diogo argues that coding agents may help us write software faster, but the software they produce still largely works the way software always has. TypeSafe is taking a different approach with Jev: putting intelligence inside software itself, so developers can build programs that reason about intent and make probabilistic decisions rather than simply generate text for a human to interpret. They discuss why reliability is the key to making AI genuinely programmable, how this could open a new era of probabilistic software, and why established SaaS companies may be particularly well positioned to benefit. Ultimately, Diogo’s goal is straightforward: technology that can reliably “do what I mean.”
Transcript
8,408 words · 4 speakers
Speaker names marked are inferred from speaking time and the show’s known hosts — they are not voice-verified.
Topics
- Jev
- OpenAI
- TypeSafe
- Codex
- Natural Language
- Claude Code
- State Machine
- SaaS Apocalypse
- Probabilistic Programming
- SVM
Page updated · transcript and summary are machine-generated