AI agents must rewrite their supporting infrastructure every 12-18 months due to paradigm shifts.

Recursive self-improvement (RSI) is now possible because the agent harness runs on the same medium, code, that the agent produces and can modify at runtime, unlike weight-based training.

“Whereas now the harness as it runs, the agent is producing and writing code and can change its own code as it runs.”

Software products in computational biology are now expected to have a useful lifespan of about one year, down from a traditional 20-year horizon, due to the pace of model improvements.

“Now it's supposed to last maybe one year.”

Reinforcement learning training time is the main wall-clock bottleneck at Poolside because batch size constraints prevent scaling by adding more GPUs.

“My biggest wall clock bottleneck right now is RL time, right? And it's just because I can't scale it up further because I can't add more GPUs to it because of that bad size constraint.”

Agents require constant LLM version upgrades, contradicting the stable deploy-and-forget nature of traditional enterprise software.

“Agents are different than normal software in that normal enterprise software, you can deploy it, put it on some server, and let it run for like 5 years, and as long as you never touch it, it just works. agents don't really work that way. The core thing that you're building on is constantly changing over time. You can't just run 4.6 for the next 3 years and just leave it. You got to upgrade to 4.8, 4.9, 5. And every one of those upgrades changes things.”

Databricks is rewriting its database engine from scratch using a decade of production workflows and billions in revenue as a knowledge base.

“Let's go back to the drawing board and design knowing everything we know today after a decade of workflows and probably billions in revenue, let's attempt to rewrite it from scratch and actually make sure it will work and it can support all these use cases.”

Decagon expects that the infrastructure needed to deploy models in the enterprise will become commoditized within three years.

“Once that gets commoditized because the agents can build that on the fly that I don't know and we'll figure out in three years from now.”

Composio's skills system enables approximately 90-95% of skills to work when switching between frontier model providers such as GPT models.

“I've tried some with GPTs like 90% of the skills just work.”