get grilled before you build (+ are you in Claude's limits test?) | LAB 0028 field notes


Lenny's AI Builders - LAB 0028

the grill-me prompt, the problem it found that i didn't know i had, and how to check if you're in Claude's limits test.

before Claude builds anything, let it grill you first. Matt Pocock's free grill-me skill makes Claude ask you questions a few at a time, each with its recommended answer, until you both agree on a plan.

i asked it for a Threadify view for the iPhone Duo's big inner screen. it found a problem i didn't know i had: the big screen was getting Threadify's cramped phone layout.

why it matters

when you build something you've never built before, the expensive mistakes are the ones you don't know to ask about. the unknown unknowns.

grill-me goes and does the research, then asks you about them before a single line gets written. you just pick its recommended answer or change it.

what you get

  • the questions you didn't think to ask. in rounds, so you're never buried.
  • a recommended answer for every question. agree in one word or push back.
  • a shared plan before any building. fewer rebuilds later.

the real example: my result

my agent ran it for me on 23 October 2026, in a fresh Claude Code session (Opus 5.5), with one line: i want a Threadify view made for the iPhone Duo's big inner screen.

  • it researched the Duo by itself first, then asked six numbered questions (Q1 to Q6), each with a recommended answer.
  • what it found: held upright and unfolded, the Duo's inner screen gives a website about 626 pixels of width. Threadify only switches to its wider layout at 768, so the big screen showed the cramped phone version. turned sideways (about 890) it jumped to the desktop layout, which doesn't fit either.
  • what it recommended: a new in-between layout for screens 600 to 900 pixels wide, not a separate "Duo mode". that also covers iPads.
  • its first build step: measure the real widths on a Duo, because Apple hasn't published them.

try it

  1. install Matt Pocock's skills: run npx skills@latest add mattpocock/skills and pick grill-me.
  2. open Claude Code in your project.
  3. type the prompt below, with the thing you want to build.
  4. answer each round. it gives you its recommended answer every time.
  5. only build once it says you've reached a shared plan.

copy this prompt

type this in Claude Code, and swap in what you want to build:

/grill-me i want a Threadify view made for the iPhone Duo's big inner screen

this is the prompt from the episode (the practical starts at 10:03). the skill is Matt Pocock's.

you'll know it worked when

it asks before it builds: numbered questions, a recommended answer on each, and at least one question about something you hadn't thought of.

a bad result: it starts writing code straight away, or asks questions with no recommended answers. check grill-me is installed and that you typed /grill-me at the start.

bonus: are you in Claude's limits test?

SemiAnalysis says Anthropic is testing new limits on some accounts: about 20% more on the five hour limit, but about 20% less over the month. great for sprints, worse if you use Claude all day, every week.

  1. in Claude, go to settings (bottom left), then the usage tab.
  2. note your weekly bar, then burn about 20 points off your five hour bar.
  3. check the weekly bar again. moved about 5 points: normal. moved 6 or 7 or more: you're probably in the test.

i checked 5 of my Max accounts across 12 days. none of them were in the test. you can't trust a limit you can't see.

what to do next

  • about to build something new: run grill-me first and answer one round before you decide anything.
  • mid-project: run it on the next feature, not the whole app.
  • limits feel worse: do the usage check before you start raging about it.

limits and dates

  • one run. 23 October 2026, by my planning agent, not screen-recorded.
  • the Duo widths are estimates worked out by others, not Apple's numbers. measure on a real device.
  • the limits test is SemiAnalysis's claim. the 5 vs 6-or-more rule of thumb comes from my own accounts and their figures, so treat it as a rough guide. checked on the Max 20x plan.

the rest of this LAB

  • Claude goes auto mode by default. i still run it with skip permissions.
  • 1Password lets agents sign in without ever seeing your password.
  • Grok Bot gets your phone apps.
  • Anthropic used Claude to prove code can't crash.
  • banks are catching AI agents in customer accounts.
  • Claude on the iPhone Duo.

sources

next step

watch LAB 0028 on YouTube - the grill-me practical starts at 10:03, the limits check at 0:14.

tried grill-me? tell me under the video what it asked that you hadn't thought of.

want the next one? get the next field notes. optional - these notes stay free to read.

Lenny’s AI Builders

i’m Lennox. i sift through AI Twitter and share one thing i tried with ChatGPT. for people making products, content and useful systems. five emails a week, with a weekly option.

Read more from Lenny’s AI Builders

Lenny's AI Builders - LAB 0045 one Claude Code setting, 36 to 38% fewer tokens in every pair of my test, the catch on cost, and a prompt that gives you 10 versions. if you build with Claude Code, one new setting can make it use fewer tokens. i tested this myself. every pair of runs used 36 to 38% fewer tokens. but there is a catch. my first run cost 83% more at api prices. the other two pairs came out 15% and 17% cheaper. why it matters the setting turns on Code Mode. Claude writes code that...

Lenny's AI Builders - LAB 0042 the one Grok Bot upgrade that matters, and the paragraph i add to claude.md for long Fable 5.2 jobs. Grok Bot 1.0 isn't new, but the launch post pulled 46 million views like it was. so don't worry, you're not missing out. the one upgrade that's actually decent got something like 40 times fewer views. and it let my bot walk into my Mac. why it matters Grok Bot has been out since August. xAI's own page calls 1.0 a major update to its AI teammates. the headline...

Lenny's AI Builders - LAB 0041 the conductor paragraph, the four steps to run it, and the honest numbers from my 40-hour run. i ran my viral Fable 5.2 conductor for 40 hours on my real app. one conductor, 73 worker threads, 25 pull requests merged. and my weekly Fable limit was gone in 13 hours. so here's what it actually built, and the one paragraph that runs yours. why it matters this is the new meta. you use Fable 5.2 as the conductor and Opus 5.5 as the worker threads. Fable 5.2 is twice...