~/writings · field notes
Notes from running production systems with agents: what broke, what the failure cost, and the check that now catches it. The method these came from is free and installable at traviseric.com/claude-code.
The through-lines / three rules that earned their place
The seven ways an agent fakes the word verified
Built, deployed, works, fed, and what each one costs to prove
The paragraph a person carries out, and the cold ground under it
Principles / the long forms behind the kit
The public kit carries each of these as a short entry you can install and run. The long form, with the incident that earned it, is here, and each one links the other. Together they stand behind one argument: anyone can build anything now; knowing what to build is the job.
AI is shaped by what people could measure. Taste is what they could not. Here is the structure I built to give a model both: a Brain that holds the facts and a Heart that holds what moves a person. Two files you can...
Read the long formIn the kitdoc: room and groundskill: room-and-groundphilosophy: the machine is a mirror of what could be measured
A tool read months of my sessions and recommended a list of things to build. Most of them existed. The gap was never invention, it was enforcement.
Read the long formIn the kitdoc: insights
Four rungs that say what done actually means — built, deployed, works, fed — and the seven ways I have watched a check lie about which rung it was on.
Read the long formIn the kitskill: definition-of-donedoc: seven cheap liesphilosophy: the instrument that lies toward caution
Most of what an AI session knows dies when the window closes. The fix is a closeout that ends in one of two named states, with the SHAs and the loose ends written down.
Read the long formIn the kitskill: session-closeoutphilosophy: the repo already knew
Seven short entries, the reading order, and the mechanisms they produced: philosophy/README.md in the kit
Evidence in the open / field reports
An objective Engineering by Selection Pressure field report about a payment path that looked complete while settlement ownership went unverified—and the shadow control built from the failure.
Read the evidenceSix months of outreach automation ended at queues nothing consumed. Here is how we now tell a legitimate gate from a hedge — and make silent caution impossible.
Read the evidenceAll notes / 28 published
The dispatcher skill, as described in the dispatcher file, is tasked with managing the founder's agent fleet from a single machine by reading every live session and assigning tasks based on permissions. This system is...
Read the noteWhen I set out to improve the copy on my website, I turned to a practical system called Copy Arena. This system eliminates the guesswork from selecting effective headlines by testing them with real readers. The process...
Read the noteThe "signup-doors-walk" skill is an essential runbook for FortListOS, designed to walk through every signup door as a customer would experience it. This process involves 23 doors, as specified in...
Read the noteThe goal-finder tool is designed to reconcile a repository's documented intentions with its current state to derive prioritized, verifiable goals. It's triggered by phrases like "what's left to finish" or "build a goal...
Read the noteThe overnight build system for Motion Studio reveals a clear, efficient process for producing video content as code. At the heart of this process is the video-engine/src/engines/html-motion/, which ensures that each...
Read the noteIf you're working on a long, persuasive page, you know the challenge of maintaining detail and precision throughout. That's where "section-pass" comes in, a method that emphasizes clarity and specificity in every...
Read the noteMy overnight build system recently encountered a practical scenario where I had to utilize the "restart-board" skill. This tool is specifically designed to help me recover my work state after a system reboot, ensuring...
Read the noteThe overnight build system processed our photo-enhance skill, focusing on enhancing existing brand photos rather than creating new scenes. This approach relies on the branding/brand-world.json to ensure that the...
Read the noteThe "recording-intake" skill transforms a day's worth of audio recordings into actionable tasks, transcripts, and reports. Every step in this process is designed to ensure that the information extracted is both reliable...
Read the noteI recently took a practical approach from the overnight build system to verify generator outputs in a more reliable way. This approach is encapsulated in the render-read-fix skill, which focuses on reading the real...
Read the noteWhen I first started using the goal-autopilot system, I wasn't sure how the goal board and engine would come together to streamline our overnight build process. The .claude/strategy/01-execution/goals/ directory...
Read the noteCreating a paid-social video ad from a single brief is a structured process. The key lies in following a predefined sequence of steps, which ensures that nothing is left to chance. Let's break down the process, starting...
Read the noteThe photo rating bench is a straightforward yet effective tool for image evaluation, specifically designed with the founder's preferences in mind. The setup is simple: put a batch of images in front of the founder using...
Read the noteWhen dealing with documentation in software projects, ensuring that new or relocated docs are discoverable is as crucial as creating the content itself. I recently had to register a new doc and learned some practical...
Read the noteIn the world of build systems, "cold-reopen" is an essential skill. It isn't about verifying something immediately after it's built; it's about returning to a feature, task, or claim after it's been marked complete and...
Read the noteThe overnight build system's recent run highlighted the importance of handling routine tasks efficiently. The land operation in our system is not just about pushing changes; it's a meticulous process that ensures...
Read the noteIn my recent overnight build system review, I used the autopilot-scout tool to identify automation opportunities within our repository. The tool's purpose is to distinguish between recurring tasks and destination-driven...
Read the noteThe process of crafting communication that resonates with others is documented in the .claude/heart/ directory, specifically in files like 03THEEIGHTRIGHTS.md and taste.md. This structure is not just about writing; it's...
Read the noteUse this skill to design what a client's ads should SAY and OFFER before any money is spent - the offer, the entry offer/lead magnet, the funnel, the creative direction, the campaign structure, and the test plan....
Read the noteThe concept of a "seam check" comes from the seam-check skill, which is all about testing the boundary between two systems. This isn't about testing the components themselves, but the join where they interact. The...
Read the noteAI is shaped by what people could measure. Taste is what they could not. Here is the structure I built to give a model both: a Brain that holds the facts and a Heart that holds what moves a person. Two files you can...
Read the noteA tool read months of my sessions and recommended a list of things to build. Most of them existed. The gap was never invention, it was enforcement.
Read the noteFour rungs that say what done actually means — built, deployed, works, fed — and the seven ways I have watched a check lie about which rung it was on.
Read the noteAn objective Engineering by Selection Pressure field report about a payment path that looked complete while settlement ownership went unverified—and the shadow control built from the failure.
Read the noteMost of what an AI session knows dies when the window closes. The fix is a closeout that ends in one of two named states, with the SHAs and the loose ends written down.
Read the noteSix months of outreach automation ended at queues nothing consumed. Here is how we now tell a legitimate gate from a hedge — and make silent caution impossible.
Read the noteAI will become an extension of individual personality, communicating as you rather than for you. The shift from generic assistants to personally calibrated agents.
Read the noteWhy generating content in parallel instead of sequentially unlocks 10x speed improvements in AI systems. Learn the architecture pattern that powers Teneo.
Read the noteTake the method, not my word for it
Every note here came from a system in production and a failure that cost something. The skills and commit guards they produced are free, MIT, and installable in one command.