Charles Zuo @CharlesLeft

@dai_jiajing ⚠️ Waymo Pro Required

Just co-hosted a private event at Emergences Labs with 15 GTM leaders during the @Techweek_ . One thing keeps echoing in my mind. The roles in the GTM teams are being redesigned in ways nobody expected. It's not that AI is creating new job titles, but that the roles that kept their titles changed completely underneath. The "Marketing Manager" in 2026 doesn't do what the 2023 Marketing Manager did. The "Search Marketing Strategist" optimizes for ChatGPT and Perplexity citations, not Google rankings. Same title. Same desk. Different tasks and skills. What's a role on your team that's quietly become a different job?

keyboard_arrow_left Previous keyboard_arrow_right Next

@EmergencesLabs It’s always amazing to look back at history and see how it repeats itself. There's a certain beauty in finding inspiration there.

In 1860, people began disciplining their bodies to stand still for the camera. In 2026, we are posing our cognition for AI - restructuring our thoughts, resumes, and workflows for the algorithm below our own conscious awareness. AI doesn't need to get smarter to replace us; it just needs us to keep voluntarily codifying our judgment and handing it over.

Kolento Hou @KolentoH

a month ago

x.com/i/article/2059…

0 2 11 25K 4

You can't do this to me @claudeai

Building 10 Products in 2026: Product #3, Day 2. Everyone says AI has been changing jobs since ChatGPT launched in 2022. I built a 1.18M job description dataset to check. The data says they're wrong. 2021 → 2024: AI mentioned in 1.6% → 2.1% of job posts. Almost flat. 2024 → 2026: 2.1% → 29.1%. Jobs mentioning AI only started surging in mid-2024, not 2022.

Building 10 Products in 2026: Product 3/10, Day 1. I stitched together 1.18M job descriptions from 2021 to 2026 to answer one question: how have jobs evolved since the rise of AI? However, most job posts vanish within weeks once filled or closed. Because no universal archive exists, historical US hiring data is notoriously hard to track. So here's the workflow I used to built the dataset using Claude Code: Step 1: I asked Claude Code to research what job data already exists. It found 3 public datasets on Kaggle + Hacker News archives going back to 2021. Step 2: We read through those JDs together and defined the criteria — 102 occupations, 4 AI tiers: Native, Enhanced, Adjacent, Unrelated. Step 3: I told Claude to scale it. It spun up parallel agents scraping LinkedIn, Indeed, Greenhouse, Ashby overnight — 1,700+ search combinations. I woke up to 42K fresh 2026 job descriptions classified and ready. Total: 1.18M JDs across 5 years. Tomorrow: what the data actually means.

gemini.google.com/share/b2e5afa6…

I built "Count Us As One" for Google I/O 2026's Code the Countdown. 200 pixel creatures wander a meadow, babble past each other, scatter when you move your cursor — alone they're just noise. Every 15 seconds, the throng emerges into a single shape: a number. They squat, open their mouths in a wave, and chant from 10 down to 1. The babble becomes a chord. The drift becomes a message. That is the piece. For these creatures, a number is the only thing they can say — and they can only say it together. Noise → harmony. Random action → meaning. Inspired by Black Mirror's Thronglets. Vibe-coded in Gemini Canvas. One HTML file, pure canvas + Web Audio. Link in replies.

Joined the private event with OpenAI's Chief People Officer. Two ideas from the room, proof-of-work hiring + people becoming agent managers, finally clicked. Companies are systems too. AI fluency isn't a person property. It scales. Almost no one evaluates companies this way.

keyboard_arrow_left Previous keyboard_arrow_right Next

The workflow that made it possible: → Cowork did 3 hours of research (scraped job boards, geocoded 4,682 offices) → I wrote a detailed PRD — this was the highest-leverage thing I did all day → Handed it to Claude Code. It pushed back on my tech stack before writing a single line → I iterated by dropping screenshots of bugs. 30-second fix cycle. The PRD > the code. Write the spec first.

Gave myself 8 hours to build something I've been thinking about for months. I can't write code. Never have. But I wanted to see: can a non-technical PM actually ship a real product in one day using only AI? 15,352 AI jobs. 1,802 companies. 41 countries. One 3D globe. Zero lines of code by me. Here's exactly how it went 🧵👇

@APompliano Both true. Tom's Hardware Q1 2026 data: 78,557 tech jobs cut, 47.9% AI-attributed. But: AI roles up 88% with $206K avg salary (Andrew Ng). IBM tripling entry-level hires. Salesforce hiring 1,000 grads. The story is reallocation, not net destruction.

@eng_khairallah1 Jensen's exact distinction is what Zapier's current AI fluency hiring standard looks like. 'Capable' = uses AI; 'Adoptive' = orchestrates AI; 'Transformative' = re-engineers workflows. The bar isn't AI usage anymore — it's stack fluency. That's what 'AI-native' actually means.

@Trace_Cohen @mcuban Agreed! Take hiring as an example: most startups are now hiring for AI-enhanced roles, traditional Engineer, PM, or Marketer positions that require AI skills. However, I am seeing startups hiring for AI-native roles, where the core tasks and skillss are built entirely around AI.

@JonhernandezIA The theory mentioned in the recent paper AI Layoff Trap (arxiv.org/html/2603.2061…) has similar views. If the market cannot create new jobs as fast as old ones disappear, we might get caught in a 'less jobs -> low consumption -> even fewer jobs' death loop.

@BSCNews The Anthropic data (anthropic.com/research/labor…) is even more counterintuitive: most-exposed quintile earns 47% more, is 4x more likely to hold a grad degree, is 16pp more female. If the displacement is real, it's a high-credential, female-skewed white-collar recession.