Aug 1, 2025 · 31m · we-live-to-build
Consider You Might Be Wrong: His Tattoo Runs His Company
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
Host Sean Weisbrot and Lotic AI CEO William Welser IV explore the critical need for cognitive responsibility in the age of artificial intelligence, advocating for intellectual humility, intentional tool use, and relational systems that challenge rather than flatter human users.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Sean holds 22.6% of the talking time here. How this is scored →
Sean as informed peer 3.5 Guest teaching 4.1 Guest disagreement 1.9 Sean pushing back 2.1
speaking balance: gold is Sean, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 5:03 Welser rejects LLM capability in emergent human nuance
Welser firmly outlines the limits of LLMs, arguing they are purely rule-based against existing literature and completely unable to handle emergent human dynamics.
Hardest push from Sean ▶ 15:55 Sean challenges optimistic adoption narrativesSean firmly rejects optimistic assumptions about human discipline, arguing that smartphone history proves people will inevitably outsource their cognition to AI.
Biggest teaching moment ▶ 5:15 The missing last mile of human interactionWelser breaks down why LLMs fail at interpersonal subtleties, explaining that conversational emergence cannot be solved with static rule sets.
Sean holds their own ▶ 3:42 Sean defends empirical AI feedback utilitySean counters Welser's dismissal of LLM prompt-response dynamics by sharing his direct empirical success using transcript evaluations to improve host performance.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Sean as informed peer | Guest teaching | Guest disagreement | Sean pushing back | Why |
|---|---|---|---|---|---|---|
| The Shift from Automation to Artificial Relational Intelligence | 3 | 4 | 2 | 2 | Sean initiates the discussion on AI and behavioral science, prompting Welser to define the historical progression of AI. Welser introduces the concept of Artificial Relational Intelligence while framing mainstream LLM prompting as fundamentally flawed and sycophantic. | |
| Evaluating AI Feedback and the Missing Last Mile of Human Nuance | 5 | 5 | 3 | 4 | Sean challenges Welser's blanket skepticism toward LLM feedback by presenting his own practical methodology of using Gemini to refine sales and podcast interviewing skills. Welser acknowledges the utility while explaining why LLMs cannot replicate emergent human nuance in the 'last mile.' | |
| Intellectual Humility and the Origin of the Arm Tattoo | 4 | 4 | 2 | 3 | Welser shares the personal story behind his 'consider you might be wrong' tattoo to highlight the risk of cognitive offloading. Sean openly evaluates his reliance on AI for sensitive pricing and sales communications, which Welser compares to seeking advice from an experienced mentor. | |
| Teaching Cognitive Responsibility Through AI Assisted Learning | 3 | 3 | 1 | 1 | Welser details using Gemini to guide his son through grammatical verb tense correction without stripping personal voice or critical learning. Sean reinforces the point, emphasizing how preserving cognitive responsibility prevents intellectual dependence. | |
| Technological Convenience, Human Friction, and Treating AI as an Elemental Tool | 5 | 5 | 3 | 5 | Sean offers a pessimistic counter-argument, asserting that cognitive decline is inevitable given how thoroughly humans outsourced memory and navigation to smartphones. Welser reframes this by discussing psychological friction and viewing AI as an elemental tool like a primitive rock. | |
| Educational Modeling for AI Integration in Parenting and Workplaces | 4 | 5 | 2 | 2 | Sean asks how parents and managers can practically instill healthy AI habits in the absence of institutional guidance. Welser explains the necessity of behavioral modeling, quality control checks, and resisting unmonitored shortcuts like vibe coding at enterprise scale. | |
| The Founding Mission of Lotic AI and Data Sovereignty | 2 | 4 | 1 | 0 | Sean invites Welser to explain the founding vision behind Lotic AI. Welser delivers an extended explanation of bottom-up data sovereignty, voice-driven vulnerability, and enabling individuals to become better consumers of health and lifestyle decisions. | |
| Strategic Lessons on Navigating AI Commoditization | 2 | 3 | 1 | 0 | Sean closes by asking for the most crucial strategic lesson from five years of running an AI startup. Welser describes pivoting away from building baseline LLMs to building specialized layers on top of rapidly commoditizing foundation models. |
Statements from this episode (9)
Welser: Current LLM systems are inherently sycophantic
“So right now these systems are sycophantic. They are, they're based on our prompts. They're going to take what we say and give us back what we've asked, and then we're going to iterate.”
Welser: LLMs cannot master conversational nuance because human emergence lacks rules
“Like, all of that last mile of interaction, it's not there, because there's no rule set against it. Like, there's too much emergence that happens between an individual when you or between two individuals when you're talking together, and then there's emergenci…”
Welser: People are offloading cognitive responsibility onto LLMs
“I am worried, and I have seen lots of examples just in my
personal space where people are offloading their intellectual curiosity. They're offloading their cognitive responsibility onto these LLMs. And they're not taking it with a grain of salt. They're sayi…”
Welser: AI's true risk is humans becoming too dumb to make decisions
“I'm not really worried about AI becoming so smart that we have our, you know, AI overlords. I'm worried about us becoming so dumb That we are just reliant upon something to make any decision. We're reliant upon something to just tell us the answer.”
Welser: Humans psychologically require friction and struggle to find happiness
“From a psychological standpoint, that's kind of problematic because we require friction in our lives, which is why people will say, oh my gosh, I've got, you know, that person has everything. They have all the money. They've got all the house. They've got all …”
Welser: Software cannot scale to enterprise level through vibe coding
“Like I can build lots of things, but if I'm gonna like scale something to an enterprise level, I'm not gonna do it vibe coding. Because there's a lot of other things that are required on the, in different parts of that system to make it robust, to make it so i…”
Welser: AI algorithms can run on individual data without pooling user data
“There's more than enough data at the individual level to run these very, very powerful algorithms. I don't need to commingle my data with yours just so I can run those algorithms.”
Welser: Users will not speak vulnerably to AI without provable privacy
“Without privacy, privacy provisioning, provable security, and an ethical approach that can be shown to the individual you're never going to get them to speak that way.”
Welser: Startups succeed by treating big tech AI models as commodities
“There's been really an AI race between these large companies, which has led to a commoditization of these tools. And if you take the approach of I'm not going to fight them, I'm not going, but I'm going to learn how to ride on all of the rails. I'm going to lo…”