The Velocity Lab

Dave O'Dell and Dan McAulay

Dave O'Dell and Dan McAulay work inside engineering organizations every day, helping them ship faster with AI — not in theory, but inside their actual teams. Each week they share what they're seeing in the field: what's working, what isn't, and what most people are getting wrong. Covering Claude Code updates, AI-enabled SDLC acceleration, and personal AI agents. No hype, no BS.

Play

Open in

23 episodes
Avg 16 min
English

Episodes 23

Newest Oldest Longest

All20 Unplayed20 In progress0 Played0

Counted on this page — what you have heard stays on this device, so it is not something the list can be paged by.

#45
Friday · 8 min
I Made a Video for $0 on a MacBook Air
Dave ran AI models entirely on a 24GB MacBook Air and built a video from a script for $0. The free local stack (OpenCode, Ollama, Hugging Face), its memory limits, and where local inference already makes sense.
#44
November 2 · 14 min
My AI Assistants Started Talking to Each Other
Dave's two AI assistants started talking to each other in Slack. Dave and Dan on one assistant backed by specialist agents, and why function and access define a specialist.
#43
October 27 · 41 min
Token Cost Is the Last Problem You Need to Solve
Why does Shipwright only run on Claude Code? Dave and Dan evaluate the Agent Client Protocol, its gaps for autonomous coding, and why token cost is the last problem to solve.
#42
October 19 · 10 min
Your Agent Can't Delete What It Can't Touch
Agent security is the new DevSecOps. Dave and Dan on blast radius, zero-permission defaults, and why a sandboxed cloud agent can be safer than one running on your laptop.
#41
October 12 · 14 min
Why Everyone Needs an AI Personal Assistant
A chatbot is glorified search. An AI personal assistant is a staff member. Dave and Dan on memory, 24/7 agents, and why everyone (even skeptic Dan) should try one.
#40
September 28 · 15 min
Three Years From Now: The Tech Is Ready, Your Org Chart Isn't
A listener asks where App Vitals lands in three years. Dave and Dan say the tech is already there — org charts, committees and approval meetings are the slow part. Plus which management titles disappear, and the robots coming for physical work.
#39
September 22 · 17 min
We Haven't Written Code in Eight Months (Here's What We Do Instead)
Dave and Dan run a two-person company where billing, payroll, marketing, podcast production, and even their scheduling software run on AI agents. If two people can automate this much, what does it mean for a 200-person company?
#38
September 21 · 16 min
AI Transformation Is a Trust Problem, Not a Tool Problem
AI adoption stalls on two human prerequisites, not tooling: trusting AI and trusting your team. Dave and Dan on 8 months without hand-writing code, why bureaucracy replaces trust, and how code review became bots reviewing bots.
#36
September 15 · 13 min
The Better the Model, the Smaller the Harness
Dave and Dan dig into the harness — the orchestration layer around your AI coding agent — and why it matters more than the model. Anthropic cut 80% of Claude Code's system prompt and nothing broke. Better models, smaller harness.
#35
September 10 · 12 min
The Four Soft Skills That Decide Who Stays an Engineer
Dave and Dan make the case that coding skill is no longer the hiring bar. The four soft skills that now decide who stays an engineer: articulation, curiosity, delegation, and full lifecycle awareness.
#34
September 10 · 12 min
Why Google Can't Move Fast (And Two Guys With Six Agents Can)
DHH spent 13 months calling AI slop. Now he orchestrates agents. Dave and Dan on why companies with the best engineers and unlimited resources still can't move fast, and why the bottleneck is process, not talent.
#33
August 28 · 15 min
The Model Doesn't Matter Anymore (Here's What Does)
Open source caught up to Sonnet on paper, but Dave and Dan still won't switch. Why they default to Sonnet over Opus, what you're really paying Anthropic for, and the 0k-per-engineer model-spend rule of thumb.
#32
August 20 · 14 min
I Was Dead Set Against This Feature (Then I Shipped It)
Dave was dead set against building a Human-in-the-Loop mode for Shipwright. Then Dan shipped it overnight using agents. Now Dave's a convert. The on-ramp that eases AI-nervous teams onto autonomous coding.
#31
August 18 · 17 min
The System Beats the Model
Dave and Dan argue the harness beats the model: a disciplined system around the LLM outperforms chasing the latest, priciest model — and it's what makes AI spend and quality controllable.
#30
July 30 · 10 min
Adopt Autonomous Coding Now, Or Find a New VP of Engineering
Dave and Dan now tell every client — solo founder to Fortune 500 — to adopt autonomous coding now. Why it's a CEO call, Shipwright vs Devin, and why developers get paid to plan, not to write code.
#29
July 21 · 14 min
You're Not Getting Paid to Code Anymore (And That's Good News)
Dave and Dan make the case that developers no longer get paid to write code — and why that's actually better news than it sounds. Judgment, planning, and engineering quality still matter more than ever.
#28
July 17 · 15 min
Shipwright vs. Devin: The $26B Competitor We Accidentally Built Against
Dave and Dan compare Shipwright (their open-source autonomous coding agent) vs. Devin AI — covering cost, extensibility, deployment, and why open source wins.
#27
July 10 · 16 min
How We Accidentally Built an Open-Source Alternative to Devin
Dave and Dan tell the App Vitals origin story: the 37signals build-for-yourself philosophy, how a side project became Shipwright (their open-source Devin alternative), and why test readiness is the on-ramp to autonomous coding.
#26
June 30 · 26 min
You're Choosing Not to 10X Your Team for Free
Dave and Dan open-source Shipwright — the autonomous agent that takes you from PRD to production with almost no human intervention. 500 PRs, barely a line of code read. The pitch: refusing a free 10X of your team is a choice you'll have to justify.
#24
June 21 · 23 min
We Built an Agent That Ships 85 PRs a Week
Dave and Dan reveal the autonomous coding agent they built: talk to it in Slack, it runs a full plan-to-deploy pipeline 24/7. Last week alone it opened 85 PRs. Build the system, not the code.

Showing 1–20 of 23 episodes