HOW TAKE AI BITE LEARNED TO CATCH ITS OWN SLIPS
The session had not started yet, and the agent was already about to skip a question it was required to ask. I was resuming a project in auto mode. The previous session had ended with a light wrap-up, which is the signal that the work is meant to continue and that the next resume can take a faster path. Take AI Bite has a rule for exactly this moment: before continuing, ask the human whether to switch to the lighter resume. It is my call, not the agent’s, because it changes what the whole session does next.
Read moreIN THE LOOP, IN CHARGE OF IT, OR IN SYMBIOSIS? WHAT THE HUMAN-CENTERED TURN IN AI STILL MISSES
For a while, the most confident advice in AI coding was to stop looking. Let the agent write, trust the tests, ship faster. The team at HumanLayer went further than most: early on, they advised against reading AI-generated code at all. Then they reversed it. After ripping out and rebuilding large parts of systems that had been assembled without real review, they landed on a rule with no soft edges: read and own the code. No exceptions.
Read morePROCESS MINING FOR AI AGENTIC WORKFLOWS, PART 1: A FIELD GUIDE
Part 1 of 4 in the series Process Mining for AI Agentic Workflows. Next: Part 2, Discovery and Value. I hear the same sentence in almost every agentic AI project, and it sounds completely reasonable: “let’s put an agent on this process.” The demo that follows is usually convincing. Some time later, the production system it was supposed to become often is not there.
Read morePROCESS MINING FOR AI AGENTIC WORKFLOWS, PART 2: DISCOVERY AND VALUE
Part 2 of 4 in the series Process Mining for AI Agentic Workflows. Previous: Part 1, A Field Guide. Next: Part 3, Engineering and the Agent. Every agentic AI project I have seen starts with a sentence that sounds reasonable and hides the whole problem: “Let’s put an agent on this process.” The word doing the damage is this. Before anyone can automate a process, someone has to answer what the process actually is, and that answer is almost never the one in the handbook.
Read morePROCESS MINING FOR AI AGENTIC WORKFLOWS, PART 3: ENGINEERING AND THE AGENT
Part 3 of 4 in the series Process Mining for AI Agentic Workflows. Previous: Part 2, Discovery and Value. Next: Part 4, In Production. Part 2 closed on a warning worth repeating: the data you mined to understand a process and the data your agent runs on in production are not the same plumbing. That single confusion is where a lot of agentic projects quietly lose a quarter and a budget.
Read morePROCESS MINING FOR AI AGENTIC WORKFLOWS, PART 4: IN PRODUCTION
Part 4 of 4 in the series Process Mining for AI Agentic Workflows. Previous: Part 3, Engineering and the Agent. Start at Part 1, A Field Guide.
Read moreFROM WARM-UP TO THE HARD CLIMB: CONTRIBUTING ACCRINTM AND ACCRINT TO A RUST SPREADSHEET ENGINE
The warm-up: shipping ACCRINTM Last spring I set out to add two functions to IronCalc, an open-source spreadsheet engine written in Rust. Both compute accrued interest on bonds. One took a couple of focused sessions. The other turned into a multi-week investigation that ended in a public discussion with the maintainer about which of two Microsoft documents is “correct.” This is the story of both, and of the working method that let the hard one span weeks without losing the thread.
Read moreHOW TAKE AI BITE LEARNED TO BOOTSTRAP ITSELF
It is curious how AI, in this case Claude Code, will embark on a quest with topics not entirely explored. That would write the fate for an explorer in counted minutes. However, we humans sometimes think that an AI suggests acting on something based on training, as if training is equivalent to knowledge. Try entering the Amazon jungle without knowledge and see if you walk out at all.
Read moreRUNNING A PMO FOR AN AI-AGENT WORKFORCE
I closed my last post by saying there was a word for the role behind the methodology and that a next post would get to it. This is that post.
Read morePROJECT MANAGEMENT FOR THE AGENTIC STAKEHOLDER
I was reading my own inbox folder when it clicked. Not the email kind, the one that lives at _inbox/ inside all projects developed with Take AI Bite, where messages from one repository land for another to pick up at the next session start. I had built it months earlier to stop losing observations between projects, populated it without ceremony, processed it without ceremony, moved entries to done/ when finished. It worked. I had not given it a name beyond “the inbox,” and I had not asked where the pattern came from.
Read more- 1
- 2