Crazy Wisdom
Stewart Alsop III | AI, Consciousness & Technology
Podcast
Episodes
Listen, download, subscribe
Episode #568: AI Is Making Everything More Efficient. What Happens Next?
Stewart Alsop sits down with Juan Verhook, founder of Tender Market, for a second conversation that ranges from the mechanics of European public tenders to the future of how we organize digital information. They cover how Tender Market helps smaller companies work around barriers like SOC 2 and ISO certification requirements, the surprising scale of public procurement (roughly 20% of GDP), and how AI and machine learning are reshaping the bidding process. From there the conversation opens up into bigger territory: the changing tolerance for being wrong in an AI-saturated information landscape, how language and culture shape perception, the reverse Turing test and the challenge of verifying human versus AI identity online, and Juan's daily workflow running eight or nine MCP servers through Claude Code. They close out talking about whether the folder and file system will survive the shift to AI-native interfaces, tying back to Stewart's own Stewart Squared episodes on the history of the PC. You can visit Tender Market at tendermarket.eu.Timestamps05:00 — Tender Market's origin story and how they help smaller companies work around SOC 2 and ISO certificate barriers. 10:00 — Public procurement and its scale, roughly 20% of GDP, plus a look at public-private partnerships. 15:00 — Local LLMs on a plane with no Wi-Fi, and comparing local model performance to frontier models. 20:00 — Supply versus demand in AI infrastructure and whether hyperscaler token efficiency is quietly improving. 25:00 — Whether AI will replace knowledge work tasks, and the shifting reality of what lawyers and other professionals actually do. 30:00 — Reverse Turing test, digital identity verification, and the idea of a "pre-AI internet." 35:00 — Model poisoning, RLHF, and the difference between pretraining and post-training. 40:00 — Interleaved tool calling and how Tender Market ties pricing to task deliverables instead of billable hours. 45:00 — RAG versus fine-tuning, prompt engineering, and when context windows actually matter. 50:00 — Deterministic programming versus probabilistic agents, and when to build custom tools versus buy existing ones. 55:00 — Juan's daily MCP stack (Supabase, GitHub, Calendly, CRM), and whether the folder-and-file system will survive the shift to AI-native interfaces.Key Insights Certification requirements aren't dead ends—they're routing problems. When smaller companies got rejected from tenders for lacking SOC 2 or ISO certificates, Juan didn't turn them away. He found that EU procurement rules allow bidding as a consortium or subcontracting to a certified partner, turning a disqualifier into a workaround that builds trust with clients.Public procurement is a massive, underexamined market. Roughly 20% of GDP flows through public purchasing of private-sector goods and services, yet most people have no visibility into how tenders work or how governments post and award these contracts.Being wrong has become more socially acceptable. Juan traced this shift to the falling cost of information: in the Stack Overflow era, giving a wrong answer was costly, but now that answers are instant and abundant, both mistakes and corrections happen faster, changing how people learn and communicate.Task-based pricing beats hourly billing for AI-era services. Rather than charging per hour, Tender Market prices around the deliverable, winning a tender, which avoids the perverse incentive of hourly billing to be inefficient and instead rewards actually solving the client's problem.RAG and fine-tuning solve different problems. RAG helps a model reference large documents without hitting context limits, while fine-tuning changes a model's internal weights so it learns new behavior or style. Juan noted that true RAG use cases needing thousands of pages of context are rarer than the hype suggests.Deterministic code should replace repeated LLM calls once a pattern is found. Stewart described his own workflow: solve a task with an LLM a handful of times, then convert the repeated pattern into deterministic software so tokens are no longer spent on it, freeing the model for genuinely new problems.AI agents are never truly autonomous. Both hosts agreed that no matter how many steps an agent chains together, a human operator always initiates the first prompt, meaning accountability and intent trace back to a person even in multi-agent systems.
Crazy Wisdom RSS Feed
