Mon 23 March 2026

This made me laugh but do not want to be direspectful to neurodiversity: Here's the thing nobody tells you about working with these models. You're basically managing an employee who is, and I've thought about this a lot, an autistic savant with amnesia. Genuinely brilliant. Solves problems in 10 minutes that would take a junior dev three days. Sees edge cases you missed. Writes elegant code. And then, mid-conversation, mid-task, just... gone. Lobotomized. Doesn't know who you are, what the project is, or why you're upset

Don't laugh...engineers want to get their hands on tokens and compute...Jensen Huang, the leather-jacket-wearing CEO of Nvidia, seemed to capture everyone’s imagination when he floated the notion at the company’s annual GTC event earlier this week that engineers should receive roughly half their base salary again — in tokens. His top people, by his math, might burn through $250,000 a year in AI compute. He called it a recruiting tool and predicted it would become standard across Silicon Valley.

COGNITIVE SURRENDER! Source: "Thinking—Fast, Slow, and Artificial" by Steven D. Shaw and Gideon Nave (papers.ssrn.com) The paper argues that AI isn't just a tool. It's a third thinking system. You know Kahneman's System 1 (fast intuition) and System 2 (slow analysis)? They're saying AI is now System 3, an external cognitive system that operates outside your brain. And when you use it enough, something happens that they call Cognitive Surrender. Cognitive Surrender is when you stop verifying what the AI tells you, and you don't even realize you stopped. It's different from offloading, like using a calculator. With offloading you know the tool did the work. With surrender, your brain recodes the AI's answer as YOUR judgment. You genuinely believe you thought it through yourself. Here are the numbers from their experiment. 1,372 participants, 9,593 trials. When AI was right, 92.7% of people followed it. Fine. But when AI was WRONG, 79.8% still followed it. Almost 80% of people went with a wrong answer because AI said so. It gets worse. Without AI, people scored 45.8% on their own. With correct AI they hit 71%. But with incorrect AI they dropped to 31.5%. That's BELOW their baseline. Meaning when AI gets it wrong, you actually perform worse than if you had no AI at all.

This is important to watch this. The WSJ is saying: AI agents and vibe-coding are disrupting the SaaS market without killing it outright, creating a dual economic dynamic: downward pressure on traditional per-seat pricing (e.g., software stocks wiped out $2T in early 2026 as agents cut seat needs, per reports on Salesforce/Atlassian declines ) while expanding the overall software TAM beyond the typical 3-7% IT budget cap by automating services markets 10-100x larger (e.g., $4B legal software vs. $400B services ). Enterprises like Cisco ($5M annual savings replacing tools) and EY ($1B tech budget, dodging SAP upgrades) vibe-code custom agents atop legacy SaaS—reducing license reliance and shifting vendors toward data platforms with AI overlays—prompting pricing innovation like outcome-based models ($1-2 per resolved ticket) or modular AI add-ons, which boost ARPU 14-23% via dynamic value pricing (Zoom example ). Net impact: SaaS prices soften for commoditized features but market grows via efficiency (30% opEx cuts ), agent adoption (51% deployed, 35% planning; 70% coding tools by 2028 [prior Gartner]), and new workflows—favoring adaptive incumbents while commoditizing rigid per-seat plays. Follow-ups Build an interactive dashboard tracking SaaS market cap, AI agent adoption rates (51% deployed), and pricing model shifts with filters for sectors like legal and IT services Computer What new SaaS pricing models replace per-seat licensing How will AI agents adoption timelines vary by company size Examples of companies shifting to outcome-based pricing What ROI metrics from McKinsey on AI sales agents

Really interesting: Sophisticated AI users at a company show wide role variation but share four key behaviors: (1) they push boundaries with longer, iterative prompts, frequent use, and model-switching for ambition; (2) treat AI as a reasoning partner via role-playing, examples, self-checks, and structured thinking to refine outputs; (3) delegate complex multi-step tasks with clear goals, constraints, and formats; and (4) view AI broadly as a cognitive tool for ideation, analysis, and problem-solving across scenarios, often casually. Surprisingly, top users skew above manager level—challenging assumptions that juniors adopt tools more readily—highlighting a gap between comfort and true sophistication in wielding AI as a dynamic collaborator. Really important from HBR: Sophisticated AI users at a company show wide role variation but share four key behaviors: they push boundaries with longer, iterative prompts, frequent use, and model-switching for ambition; treat AI as a reasoning partner via role-playing, examples, self-checks, and structured thinking to refine outputs; delegate complex multi-step tasks with clear goals, constraints, and formats; and view AI broadly as a cognitive tool for ideation, analysis, and problem-solving across scenarios, often casually. Surprisingly, top users skew above manager level—challenging assumptions that juniors adopt tools more readily—highlighting a gap between comfort and true sophistication in wielding AI as a dynamic collaborator.