Building And Structuring An AI Native Company
YC built an English-to-SQL agent over data on 7,000 companies and 20,000 founders.
7,000companies in YC's internal data queried by its English-to-SQL agent (with 20,000 founders)
Lecture
Week 33 · 10-16 Aug
2 sources filed · EN 2
B C Claude runs my entire business
Skills behind the pipeline: a proposal maker (versioned, with a change log), a voice guide, and an hourly operator that checks which clients need the next step.
Week 32 · 3-9 Aug
1 source filed · EN 1
B How To Design In The Agent Era
Rules against generated-design tells: pull font weights back from bold, use at most three font sizes, check contrast, delete decorative numbers, icons and badges, avoid all-caps kickers with wide letter spacing, purple gradients, coloured card edges, and cards everywhere.
threemaximum number of font sizes the speaker recommends to make generated layouts look deliberate
Week 31 · 27 Jul-2 Aug
2 sources filed · EN 2
B Jeff Dean: The 1% Rule for Building in AI
Specialised low-latency, low-energy inference hardware is his candidate for the next "fits in memory" moment; he asks what becomes possible with 50x lower latency.
0% or 1%model success rate Dean says marks a good startup problem, as opposed to 20%, which signals the general model will soon catch up
B Boris Cherny: We Cut 80% of Claude Code’s Prompt
Claude Code's harness code is now mostly safety, permissions, static analysis and UI; much other code has been removed.
80%share of Claude Code's system prompt deleted for Opus 5
Week 30 · 20-26 Jul
1 source filed · EN 1
B The AI content machine that turns ideas into posts that don't sound like slop | Alex Lieberman
Process mapping first: he wrote out his manual content process (idea, research, brain dump, format, draft, edit, publish, repurpose) and decided where AI drives and where it assists.
15content ideas ('spikes') the pipeline proposes each day, half from internal systems and half from followed accounts
Week 28 · 6-12 Jul
1 source filed · EN 1
B Why Trust Is the Bottleneck for Agents — Jean-Denis Greze, Town | AI Engineer World's Fair 2026
Town pivoted from AI for tax preparers after noticing their email and calendar overhead; the prototype took two weeks and got 20 DAUs within days.
10prior approved actions Town uses before asking a user to let an automation run without approval
Week 25 · 15-21 Jun
1 source filed · EN 1
B The AI Skills Nobody is Teaching (And Everyone Needs) | AI Expert Ethan Mollick
Prompt-engineering tricks (roles, bribes, "think hard") no longer matter much with current models; instructing it as you would a person is enough.
48% to 84%share of GDPval tasks where the best AI ties or beats experts, a year ago versus at recording, as cited by Mollick
Week 23 · 1-7 Jun
3 sources filed · EN 3
B How to Build an AI-Native Services Company
Good markets share four traits: work already outsourced (low trust, so no behaviour change), low judgment per step, a high overall intelligence threshold, and regulation that raises the bar.
B Building an AI Native Engineering Org - Angie Jones
Maturity stages: 0 no AI; 1 autocomplete; 2 chat without PRs; 3 delegating and reviewing; 4 parallel agents; 5 delegating complete tasks with shippable results.
90%share of Block engineers using coding agents after a few months, while features did not ship faster
C How to build an iPhone app with zero technical skills
First prototype came from a one-line prompt in Lovable and Replit; she kept Replit partly because a friend there could unblock her.
25 or 30 hourstime spent in one weekend moving the app from Replit to App Store submission with Claude and Claude Code
Week 22 · 25-31 May
2 sources filed · EN 2
B OpenAI @ Replay 2026 | How OpenAI Uses Codex to Change How We Build
Context: Codex is connected to Linear, Sentry, CI, Figma, Gmail, calendar, Slack and Notion, so "vague prompts" (find the thing a colleague asked me for) can be resolved.
A Inside the 2026 AI Index Report
Adoption: 88% of organisations use AI in at least one function; 53% population-level adoption of generative AI tools; global corporate investment more than doubled year on year.
88%organisations globally saying they use AI in at least one business function
Week 21 · 18-24 May
1 source filed · EN 1
B How to Build a Self-Improving Company with AI
Loop structure: sensors (emails, tickets, cancellations, telemetry), a policy layer (what needs human permission, what must be logged), deterministic tools, a quality gate (evals, safety checks, human review for high-risk cases), and a learning step that feeds back.
about 5xmore revenue per employee at YC demo day than 18 months earlier, per the speaker
Week 18 · 27 Apr-3 May
2 sources filed · EN 2
B Andrej Karpathy: From Vibe Coding to Agentic Engineering w/ Stephanie Zhan
Software 3.0: programming becomes prompting, and the context window is the lever.
more than 10xspeed-up he sees for the best agentic engineers, above the old '10x engineer' notion
B Aaron Levie on AI Adoption and Enterprise Workflows | The a16z Show
Failure pattern (Casado): the board asks for AI, the CEO hires a consultant, a central project starts that nobody understands, operations are not aligned, and it fails.
80 to 90%share of a new Box feature that AI built, per Levie
Week 16 · 13-19 Apr
1 source filed · EN 1
B Harness Engineering: How to Build Software When Humans Steer, Agents Execute — Ryan Lopopolo, OpenAI
Workflow: start from a ticket, hand it to Codex with a few skills that teach it to launch the app, the local observability stack and Chrome DevTools.
750packages in the team's pnpm workspace after restructuring the repo for agents
Week 5 · 26 Jan-1 Feb
1 source filed · EN 1
B AI Won’t Replace Workers. It Will Redesign Work | Andrew Ng
Ng finds the AGI label misleading.
about 5%share of staff who are very heavy, innovative AI users in companies the interviewer's firm has instrumented