Key Takeaways
- AI agents are delivering tangible results: Codex generated a perfect 3D-printable lid from a photo, and Reflection Engine produced startling personal insights.
- Security best practices are shifting: Storing production .env files in repos is now risky because AI agents may unknowingly access them.
- Startup growth indicators are strong: Supabase adds over 1 million Postgres databases weekly, and Podscan now ships TB-scale podcast data to financial firms.
- Pricing changes are under scrutiny: Mintlify's subscription price increased from $340 to $540/month over three years—a 3x jump some argue isn't justified.
- Compute funding remains fragile: A European VC withdrawal cost Ben Tossell about $500K in compute credits, exposing infrastructure deal risks.
1. AI in Practice: Capabilities, Risks, and Personal Reflections
- Codex demonstrated a tangible win: @dvassallo shared how Codex helped someone identify the dimensions of a lost coffee grinder lid and generate a perfect 3D-print replacement file. He noted he's had similar experiences and remains impressed by the capability. — via 1
- Ben Tossell published results from Reflection Engine, an AI-generated personal behavior report, despite feeling unprepared to read it. Key takeaways included "unemployable but don't apologize" and a realization that he doesn't want to run an education business, prompting him to recommend the tool. — via 1 2
- Arvid Kahl warns that putting production .env files in code repositories is now extremely dangerous, because AI agents may not know which environment variables their called tools will ultimately use. He strongly recommends using development credentials only to avoid leaks. — via 1
2. Startup Growth, Pricing, and Business Models
- Supabase continues to scale rapidly: @bentossell highlighted that Supabase adds over 1 million Postgres databases per week (excluding read replicas), underlining the platform's explosive adoption. — via 1
- Mintlify's subscription price has risen from $340 to $540 per month, a 3x increase over three years. @JackEllis argues that while the product is good, the improvements don't justify the steep hike, sparking a broader debate about SaaS pricing. — via 1
- Podscan has evolved from a podcast keyword monitoring tool into a provider of TB-scale podcast data to large financial intelligence firms and research labs, powering prediction and analysis engines. @arvidkahl described the journey as "a wild ride," highlighting the growing value of voice data. — via 1
- @tibo_maker shared a tool stack that supports a $1M MRR business across AI coding, video content, SEO, payments, and affiliate marketing, with total costs below a junior employee's weekly salary. He used this to encourage founders to start building. — via 1
- Commenting on Airtable's acquisition, @arvidkahl criticized those who scoff at it, asserting they have no idea how difficult it is to build a company. This ties into broader sentiments about founder struggles and exit realities. — via 1
- @yongfook proposed that a business can be built on painkillers or vitamins for a large or niche audience, but "vitamins + niche audience" is not a viable combination. This framework offers a quick litmus test for startup ideas. — via 1
3. Product and Infrastructure Challenges
- Mole, a Mac cleanup tool created by @HiTw93, received a positive review from Andy Hutchinson of Mostly Mac, who cleared about 13GB on his M2 and called it "no subscription, no bloat, no hassle" after six months of daily use. Tw93 was delighted, noting the project began as a weekend shell script. — via 1
- @bentossell revealed that a European VC backed out after signing a term sheet, wasting about $500K in compute credits. He is now publicly asking for large compute resources, underscoring the fragility of AI infrastructure deals. — via 1 2
