Peter has been reading the research on why organizations never throw anything away: subtraction neglect, planned abandonment, structural inertia. Ed has a consulting gig whose whole remit is "tell us what to throw away" after nine months of AI adoption and a C-suite split down the middle. They disagree about whether cheap inference will last, and agree that vendor lock-in hurts a hundred-person team far more than an indie. Then Peter shows off AirDash, an Airtable for agents built on Cloudflare's per-tenant SQLite databases: agents read with SQL, write through MCP, and the human gets a friendly table view.
Ed's Scalpels launch week had no marketing at all, because a client bought a week of his time. Their fifteen-year-old, pre-Composer CRM runs the business, and an AI-built replacement came out no faster than the original because its requirements guarded against every bad day since 2007. Ed started over: two and a half days triaging requirements, continuous benchmarking at five times the target throughput against an SES simulator with dials for latency and failure, tombstoning to find the code paths that are actually live, and no reporting dashboard at all, because a Claude plugin with an MCP and skills lets each stakeholder build the reports they want. That has him wondering why Scalpels has a dashboard at all.
In the second half they talk marketing. Peter automated his SEO and social loops for Sputnik and worries he should have sold first; Ed pushes back. Peter argues that Ed's own stories about running a client fleet with Scalpels are the marketing, and Ed lays out the plan: one Claude plugin, one OAuth, every client team visible. They finish on what's left for indie founders when building and maintaining software are both fixable: attention, data, hard problems, and selling outcomes instead of tools. Plus why "my agent did it" is no excuse, and rivers that can sue.
Chapters
00:00 Autumn in Basel, winter sun in Madrid
01:30 Every company is figuring out AI
02:53 Keep a loose grip on your experiments
04:02 Subtraction neglect and planned abandonment
04:47 A consulting gig: tell us what to throw away
07:19 Healthy skepticism and subsidized inference
11:30 Vendor lock-in hits big teams hardest
13:55 Peter builds AirDash
16:11 A million SQLite databases for five bucks
18:45 When the Mac mini runs out of room
21:43 SQLite as an exit strategy
23:04 Launch week with zero marketing
23:13 A fifteen-year-old CRM
25:10 Requirements written from past pains
27:15 Starting over, with a week to do it
28:49 Benchmark before you write code
30:15 An SES simulator for load testing
34:47 Why this takes a senior engineer
38:08 Sunk cost features
39:11 Tombstoning a pre-Composer codebase
42:51 Gutting the reporting dashboard
44:25 A Claude plugin instead of a dashboard
49:26 Client work pays the bills, and the first paying user
51:37 Peter automated marketing before selling
53:53 Plant seeds now
57:29 Your stories are your marketing
59:11 One plugin for your whole client fleet
1:01:24 Monetizing without a paywall
1:03:35 Why launch deadlines matter
1:03:57 Peter's two-products-a-year rule
1:06:12 Is maintenance hell fixed?
1:07:57 What's the moat for indie founders?
1:09:51 MCP points and artificial limits
1:11:23 AI makes attention harder
1:12:04 Selling outcomes
1:14:58 Solve something hard, sell outcomes, or sell data
1:17:28 "My agent did it" is no excuse
1:18:47 Rivers with legal rights
Links
Artisan Beer: https://artisan.beer
Scalpels: https://scalpels.app
Hone: https://scalpels.app/products/hone
Our CVEs: https://ourcves.com
Sputnik Intelligence: https://sputnikintelligence.com
Ed / Artisan Build: https://artisan.build
Peter: https://petervandijck.com
AirDash: https://getairdash.com
Mentioned
Peter's post, How do organizations throw (AI) stuff away?: https://petervandijck.com/post/hr0bn-yblas
People systematically overlook subtractive changes (Nature): https://www.nature.com/articles/s41586-021-03380-y
Conway's law: https://en.wikipedia.org/wiki/Conway%27s_law
Airtable: https://www.airtable.com
Bending Spoons: https://www.bendingspoons.com
Cloudflare Durable Objects SQLite storage: https://developers.cloudflare.com/durable-objects/api/sqlite-storage-api/
Amp: https://ampcode.com
Solo: https://soloterm.com
Amazon SES: https://aws.amazon.com/ses/
PHPStan: https://phpstan.org
Claude plugins: https://claude.com/plugins
MCP Apps: https://blog.modelcontextprotocol.io/posts/2025-11-21-mcp-apps/
Aaron Francis: https://aaronfrancis.com
NativePHP: https://nativephp.com
Laravel Cloud: https://laravel.com/cloud
Pieter Levels: https://levels.io
Bento: https://bentonow.com
Stripe Atlas: https://stripe.com/atlas
Railway: https://railway.com
If you enjoyed it, like and subscribe. Peter still insists.
#Laravel #AI #MCP #IndieHackers #SaaS #Podcast