Back
20/65
Lesson 20
12 min
Case Study: SLB
From tweet to working tool in one evening
Ready to level up?
Mark complete to track your learning progress.
From tweet to working tool in one evening
Mark complete to track your learning progress.
See how a tweet becomes working code in one evening: 76 beads, 268 commits, from idea to ~70% complete in hours.
On December 13, 2025, a conversation on X about AI agents accidentally deleting Kubernetes nodes sparked an idea: what if dangerous commands required peer review from another agent?
"You know how in movies like WarGames they show how the two guys have to turn the keys at the same time to arm the nuclear warheads? I want to make something like that where for potentially damaging commands, the agents have to get one other agent to agree with their reasoning and sign off on the command."
The idea was simple: like the "two-person rule" for nuclear launch codes, agents should need a second opinion before running destructive commands like rm -rf, kubectl delete, or DROP TABLE.
Instead of just noting the idea for later, the flywheel approach is to start immediately while the idea is fresh.
Once the initial plan existed, it was sent to multiple frontier models for review and improvement:
Architecture refinement
Edge case analysis
Security considerations
Combining all feedback
The feedback was then integrated by Claude Code, with multiple verification passes to ensure nothing was missed:
1# First pass: Integrate all feedback2cc "Revise the plan document using all the feedback.3Make sure ALL changes are reflected properly."45# Second pass: Verification6cc "Go over everything again. Did we miss anything?"7# Result: Found small oversights89# Third pass: Final check10cc "One more careful review. Any remaining gaps?"11# Result: Found 2 more edge cases
The refined plan was then transformed into structured, trackable beads. The prompt was carefully crafted to ensure thoroughness:
1cc "First read ALL of the AGENTS.md file and2PLAN_TO_MAKE_SLB.md file super carefully.3Understand ALL of both! Use /effort max.45Take ALL of that and elaborate on it more, then create6a comprehensive and granular set of beads with:7- Tasks and subtasks8- Dependency structure9- Detailed comments making everything self-contained10- Background, reasoning, justification11- Anything our 'future self' would need to know1213Use the br tool repeatedly to create the actual beads."
Smaller than cass-memory's 693 beads, but still comprehensive enough to capture the full implementation.
Then, just like the plan itself, the beads went through verification passes:
1# Beads verification prompt2cc "Check over each bead super carefully:3- Does it make sense?4- Is it optimal?5- Could we change anything to make the system work better?67If so, revise the beads. It's a lot easier and faster8to operate in 'plan space' before implementing!"
The Simultaneous Launch Button implements a two-person rule for AI coding agents:
System destruction, database drops
rm -rf, git push --force
Single file delete, branch remove
Temp file cleanup, cache clear
With beads ready, the agent swarm began implementation. The project was smaller than cass-memory, but the workflow was identical:
1# Launch agents2ntm spawn slb --cc=3 --cod=234# Each agent runs:5bv --robot-triage # What's ready?6br update <id> --status in_progress7# ... implement ...8br close <id>910# Commit agent runs every 15-20 min11cc "Commit all changes in logical groupings with12detailed messages. Don't edit code. Push when done."
By dinner time, about two-thirds of the project was complete. The agent swarm continued working while the developer ate, pushing commits autonomously.
Compared to the 693-bead cass-memory project, SLB's 76 beads allowed for some workflow optimizations:
An hour from idea to initial plan keeps momentum high
Each model brings different perspectives and catches different issues
Each pass found something - never skip this step
76 beads is manageable while still demonstrating the full workflow
The conversation transcripts become valuable learning resources
Pick a small tool idea (something that would take you a day or two manually) and try this workflow:
1# Hour 1: Draft initial plan2cc "I want to build [your idea]. Help me create a3detailed plan document covering architecture,4features, and implementation approach."56# Hour 2: Multi-model feedback7# Send plan to 2-3 different frontier models8# Collect their suggestions and improvements910# Hour 3: Synthesize and create beads11cc "Read the plan and all feedback. Create a12revised plan incorporating the best suggestions."1314cc "Convert the plan into 50-100 beads with15dependencies. Use br CLI."1617# Hour 4+: Implementation18ntm spawn myproject --cc=2 --cod=119# Let the swarm work!2021# Every 15-20 min: Commit agent22cc "Commit all changes with detailed messages."