AI tools, tested on my real setup.

Every week I run one AI tool on my own agent, measure it, and give it a YES, LATER or NO. The tests are free. The setups I use are for sale.

Latest: TEST 002All tests
LATER

One tool per episode.

  • TEST · A real tool on my setup, with my numbers and a verdict.
  • BUILD · Something I automated, shown end to end.
  • SETUP · A how-to from my own stack, in a few steps.
LATER

TEST 001 · TRIM

llmtrim: −16.5% tokens, still off

Read the test →

BUILD 001 · AGENT

An assistant that does my to-do list

Read the test →

LATER

TEST 002 · CODE

Code Mode: 1.17M tokens to 1K

Read the test →

SETUP 001 · VAULT

A vault that maintains itself

Read the test →

TEST 003 · FLIP

The coin-flip trading skill

Coming soon

BUILD 002 · VOICE

My agent answers in 3 seconds

Coming soon

BUILD 003 · REELS

Claude Code edits my reels now

Coming soon

llmtrimLATERAn assistant that does my to-do listLIVECode ModeLATERA vault that maintains itselfLIVEThe coin-flip trading skillSOONMy agent answers in 3 secondsSOONClaude Code edits my reels nowSOON

TEST 002 · Code Mode

The numbers, measured.

4

episodes tested so far

Reel frame: Cloudflare's README table, about 1.17M tokens vs about 1,100 for Code ModeReel frame: the program runs in a sandbox, only the answer comes backReel frame: the pattern ADOPT, their MCP server LATER

The lab desk.

The files behind TEST 002, straight from my setup. Drag them around.

Read the full test
summary.txt
Cloudflare's way of letting an agent call tools through code: what it does, their numbers, and whether I'd use it.
results.txt
verdict.md
LATER
TEST 002 · comment CODE
Free tests. Paid setups.
The write-ups stay free. When a test earns its place in my stack, the exact setup becomes a pack you can copy in an afternoon.
PACK · REELS
The reel pipeline
  • ✓The code that makes these reels from a script
  • ✓Free voice clone and lip-sync setup
  • ✓Scene templates in this look
Coming soon
Follow for the launch
PACK · TOKENS
Cut your agent's bill
  • ✓Prompt caching, set up step by step
  • ✓llmtrim in safe mode, scoped to one app
  • ✓A script that measures your savings
  • ✓A 10-point cost checklist
Coming soon
Follow for the launch
PACK · VAULT
Self-maintaining vault kit
  • ✓The two nightly scripts, add-only
  • ✓A rules file for your folders and tags
  • ✓Dry-run, cron and n8n setup
Coming soon
Follow for the launch
PACK · AGENT
Agent task desk kit
  • ✓The rating and plan prompts
  • ✓A fenced job script: own branch, spend cap, draft PR
  • ✓Limits and merge rules
Coming soon
Follow for the launch
Every test stays free.
Comment the keyword on any reel and the full test lands in your DMs.
Browse the tests