Agentic Development Gets an Adversarial Layer as Supabase Acquires Turso

Key Takeaways

1. Arvid Kahl outlines an agentic dev workflow built on adversarial review, codebase-resident context, and production-shaped test data.
2. Supabase has acquired Turso Database, with details promised at a keynote.
3. A self-funded founder runs 5 SaaS apps at $200K/month by avoiding AI-dependent products and using lifetime deals for pre-MRR cash.
4. Noah Kagan used Claude to audit his own onboarding and reports 6 sales from 19 SEO pages in 3 days.
5. Acquire.com listings show small AI and niche apps trading at modest multiples, while a forwarded take argues AI lets tiny teams run $10M companies.
6. dearCC launched as a free AI-powered job platform for Gen Z, with roadmap, mentorship, and community features.

1. Agentic Development and AI Tooling

1. Arvid Kahl summarized his agentic development process: every step should include adversarial review, ideally with a third agent as arbiter; test suites must cover unit through end-to-end with no exceptions, and TDD must be embedded in the agent workflow, while non-code stages like architecture and decisions rely more on human review. He also argued context should be maximized into the codebase — docs, ARDs, customer ICP descriptions, runbooks, and schema annotations belong inside the repo, not as attachments — and that each agent conversation and change should be tracked and attached to tickets or PRs. Development and test database contents matter more than ever: dry runs, workflow simulations, edge-case checks, and performance measurement should happen before production, and seed data must represent production shape even at smaller scale. He added that code best suited for agents may not look best to humans, positioning people as verifiers rather than proofreaders. 1 A forwarded post questioned the evidence for adversarial review's effectiveness, asking whether different models pitted against each other differ, how same-model improvement requests behave within versus across sessions, and who is doing this research. 2
2. Noah Kagan recommended founders use Claude to walk through their own signup and onboarding as a new user, then list the top 5 fixes with per-step observations; the exercise surfaced that his onboarding never asked users to create a booking type, the product's core action, and Claude provided a fix. 1 He also reported that a $20 Mac app got 6 PayPal sales in 3 days after publishing 19 SEO pages, and argued every product needs two SEO page types: competitor comparison pages and "Best [product] for [X]" guides for audiences like lawyers, doctors, ADHD users, and students. His Claude Code prompt asks for the top 5 competitors, vs and alternative pages for each, and 10 "best [product] for [audience]" guides, with honest notes on where competitors are better. 2
3. Ben Tossell forwarded that Factory automates low-leverage work, with automatic model routing removing the burden of picking the best of 10 new models each week. 1 He also said personal agents should proactively build apps and websites for users. 2 Separately, he criticized a product experience as not good enough for people who are "non-technical but not average users," arguing the problem should be fixed in the product rather than blamed on user skill. 3

2. Infrastructure and M&A

1. Ben Tossell forwarded that Supabase has acquired Turso Database, calling himself a long-time supporter of the Turso team and product and saying the two sides have major plans to disclose at a keynote hours later. 1 He also mentioned attending Supabase Select and noted a PlanetScale truck outside the event did not stay long. 2 3
2. Acquire.com listed a mobile AI learning platform that turns notes and media into personalized learning material, with $1.3M TTM revenue and a $1.8M asking price. 1 Other new listings include a UK driving-theory test app at $17.1K TTM revenue, $11.6K profit, and a $29.2K asking price; a jewelry e-commerce brand with registered IP, production capability, and inventory at $200K TTM revenue and a $199.9K asking price; and a Spanish food marketplace with 15,000+ users, nearly 5,000 products, over €185K annual revenue, $209.5K TTM revenue, and a $523.8K asking price. 2 3 4 A forwarded take argued AI gives small teams huge leverage, predicting more near-$10M companies with almost no middle management — for example 15 employees all close to the actual work, less bureaucracy and fewer meetings, and more money flowing to value creators — making traditional startup org charts look outdated. 5 Acquire.com also argued running a business and being acquired are different skills, and founders can use M&A advisors rather than figuring out the latter alone. 6

3. Founder Playbooks and Market Signals

1. Starter Story profiled Australian founder Mike, who runs 5 SaaS apps at a combined $200K/month, fully bootstrapped with no VC and no exits, under the principle of "only doing ideas that cannot fail." His rules include choosing validated, mature ideas over new concepts; fixing 4 co-founders at 25% each to reduce founder conflict; staffing with frontend, backend, a designer, and one all-round operator; never giving away free accounts because paying users actually use the product; and distributing profits to founders once growth reaches $10K MRR to cover costs, staying lean long-term. Before MRR, he funded via lifetime deals: one-time pricing of $59–100, with a private LTD for Thrill alone raising $30K, then AppSumo, then a final "last chance" LTD permanently closed, targeting $100K cash before relying on MRR. He refuses to build any AI-dependent product, saying "you can't control it, the risk is huge," and is watching boring documentation tools, which he considers overpriced and underfeatured, noting AI needs quality documentation to recommend your product. His advice to his younger self: work with people you'd have a beer with. 1
2. dearCC launched as a free platform for Gen Z job seekers, offering a roadmap and network, anchored by @SchultzFamilyFd and @TheBGInstitute. It provides three resources: an AI-driven labor intelligence Field Report to identify growing and stagnant roles; a personalized step-by-step Game Plan to build AI capability and run a targeted job search; and a community accountability mechanism of industry mentors and peer Crews. The forwarded post described the current graduate job market as one of the hardest in decades, arguing that sending 500 applications with no reply should not be treated as a rite of passage but as system failure, and that AI has changed the rules of job hunting. 1
3. Alex Lieberman forwarded that the real winner of the AI race is capitalism, citing lab delivery speed, downward pressure on model prices, historic infrastructure buildout, and money flowing into venture capital. 1 He also said demand for his executive AI Agent bootcamp was strong, with 200+ companies on the waitlist 18 hours after launch; it targets VP level and above at companies with $100M+ revenue, offers a free 5-day course, and prioritizes applications. 2 He cited a demo where someone tried to build personal CAD software that automatically constructs CAD models because a friend in CAD thought AI wasn't good enough yet, framing it as a limitation of CAD tools rather than model capability, and commented "AI will eat CAD." 3
4. Nick Huber argued startups require people to work for you, partner with you, and buy from you, so every problem ultimately comes down to people — someone not doing what they promised, or expectations not matching reality. Dealing with people is an all-day, every-day part of being a founder, and like sales it is not innate: you must accept discomfort and practice as if you were an extrovert, and he said he is still learning and will keep making mistakes. He argued mastering the art of dealing with people gives you the ability to build a company and become very wealthy, and once you can recruit, hire, and manage a great team, nothing can stop you. 1
5. Codie Sanchez cited a wealthy mentor's line that by the time most people finish analyzing, the other side has made 3 mistakes and found a better way, so the cost of moving slowly exceeds the cost of being wrong. 1 She also highlighted a Princeton economist and co-author's study using anonymous IRS records on how the US top 1% make money, finding that for every 1 person on the Forbes list there are over 4,000 private business owners worth $10M+. 2 Her pricing advice: don't price against competitors, since market price is just the average of other people's fear; go premium and charge more, provided you are worth it. 3
6. Alex Hormozi argued you don't need to know something will succeed before starting, only that it is better than doing nothing. 1 He announced the Scale or Fail finale is live, with 8 entrepreneurs competing for $100K, exclusively on YouTube. 2