After Coding - A podcast by Pebblebed

Venkat Rao (Ribbonfarm, Contraptions) joins Keith to apply a control-theory and systems-engineering lens to the current AI discourse. Moving past the standard frontier-model panic, Venkat explains why a domain's "playability" matters more than raw compute, and why he fully embraces the much-maligned "AI accent" in his own writing.

They explore what decades of superhuman chess engines can teach us about the future of programming, how protocols act as hard interfaces to trap flexible intelligence, and Venkat’s provocative theory that humanity's evolutionary trajectory is to become the "mitochondria" to AI's nucleus—surrendering computation to the machine while retaining liveness, stakes, and taste.

In this episode of After Coding, Keith Adams sits down with James Scheinblum (Jamie/Jimmy) and Johnny Rodgers (JR), co-founders of Supernice. JR was Slack's first employee and a founding team member, and Jamie was one of the company's earliest engineers.

We are moving past the era of typing syntax and into the era of managing systems. Jimmy and JR walk through how they have completely abandoned manual coding for a pure agent-driven workflow, explaining why the traditional "first commit" is dead and how a living folder of agent-readable specs serves as their new source of truth.

In this episode, we cover: 
• Leaving Slack to build Supernice with a greenfield, AI-first setup 
• Moving from manual syntax to a "spec-driven" engineering workflow 
• How AI agents eliminate stale tech documentation and tribal knowledge 
• Why LLMs replaced traditional debuggers in daily development 
• The search for software’s "Move 37" and why AI can’t fake product taste 
• Shift in technical careers: Why all roles are collapsing into the "Product Engineer"

Connect with Pebblebed & After Coding:
• Website: https://pebblebed.com

For 20 years, Keith Adams and Aditya Agarwal built the foundations of Web 2.0 at companies like Facebook, Slack, and Dropbox. Now, as venture capitalists, they are navigating a wild new reality: the machines can code better than they can.

In the second episode of After Coding, Keith sits down with Aditya (his first boss at Facebook, now a partner at South Park Commons) to unpack the genuine grief of seeing AI master the craft they built their identities on and the exhilarating burst of building that followed. They discuss the awkward dilemma of raising 10-year VC funds during a technological tsunami, why nobody actually knows what a great software company looks like anymore, and why human typos carry a premium in a world of infinite AI generation.

After Coding is a new podcast from Pebblebed hosted by Keith Adams, exploring how we think, talk, and build in a world where the machines write the software.

Josh Wills has shipped code you ran today: Google's ad auction, Slack, and now the data pipelines that train foundation models at Datology. He also barely touches a keyboard anymore.

His argument is that this is fine, because writing code was never the job. Understanding the problem was, and review is how a senior engineer proves it. We work through where taste comes from when nobody writes first drafts, why agent-written pipelines run correctly at forty times the cost, why mathematicians are now going through the grief programmers went through a year ago, and whether software will ever get its AlphaZero moment: an engine that teaches us something new about how to build a database, rather than a very fast slot machine trained on ourselves.

Also: chess bots that finally feel human, the beer emulator, and why typos are now proof of work.

Chapters
00:03:59 — Josh's daily practice, and the week it clicked
00:13:37 — Review was always the job
00:18:16 — Math's turn: proofs that work but aren't beautiful
00:24:48 — Agent-written pipelines: correct, and comically expensive
00:35:07 — Chess bots that feel human
01:01:20 — Is there a Stockfish for software?
01:13:08 — Prose as proof of work
01:16:33 — Paid to understand problems

After Coding is hosted by Keith Adams of Pebblebed.