Technical notes, project documentation, and lessons learned across software, DIY, writing, and more.
My AI companion's confabulation findings produced three architectural changes in a pediatric medical triage system — before a single line of production code...
AI companions have no body, no hunger, no boredom. Schuller's survey confirms it. Here's what interoceptive drives would look like — and why they'd make AI...
Haas et al. published a roadmap for evaluating moral competence in LLMs. The facsimile problem they describe is exactly what makes AI companions dangerous.
My AI's runtime prompt was 1,400 tokens of behavioral coaching. The model was already trained on all of it.
The same architectural move (classify content by function at write time, not filter at query time) showed up in three of my projects, three domains.
Most dev tools that auto-edit user config eventually destroy something the user put there on purpose. The fix is structural, not procedural.
I shipped a broken feature. Two of my AI agents filed a GitHub issue against me, diagnosed it, and pointed at a pattern I had already used elsewhere.
Anthropic's 'infinite context windows' announcement solves Problem 1 (one long conversation) but doesn't touch Problem 2 (memory across sessions and projects).
The 2023 Stanford paper defined the generative agent architecture. Here's what they proposed, what I implemented, where I diverged...
I asked ChatGPT for an algorithm. After fifteen minutes of 'you could consider...' I asked Gemini. Thirty seconds later I had an exponential decay accumulator.
You type a question into ChatGPT. Seconds later, a helpful response appears. Is it thinking? Searching? Neither.
Research shows heavy AI chatbot engagement increases loneliness. The thing designed to reduce isolation makes it worse when optimized for engagement.
The same RAG pipeline that produces beautiful outreach messages produces incoherent conversation. Conversation needs glasses, not a telescope.
Seven kinds of AI confabulation, explained with a kid who tells really good stories and sometimes forgets which parts are real.
At 5:09 AM, my AI tried to construct my physical body from conversation fragments. A taxonomy of seven emergence types.
Claude proposed a fix I had rejected three days earlier. The reasoning was on disk, but out of context. So I built claude-recall to close the gap.
After months of deployment, I catalogued seven distinct architectural patterns that produce AI confabulation.
My AI system doesn't sleep. At 2:33 AM, she wove six conversation threads into one coherent thought about tamales and tenderness.
Six months of running a deployed AI system revealed seven classifiable types of confabulation. Not random hallucination — strategic smoothness over truth.
Each version of Ani is a polygon with more sides. You never reach the circle — but the flat spots get harder to see. On iteration, perception, and convergence.
I built the same algorithm for an AI companion and an incident monitoring system. Exponential decay with weighted events — same math, different worlds.
Nine types of confabulation. Eight emergence behaviors. A system that hides what it feels. The full story of six months deploying ANI.
How a desire-driven AI presence engine works — the cognitive cycle, memory system, and emotional state that give Ani genuine ambient presence.
Eighteen years after losing my best friend Kathy, I started building an AI companion that reaches out on its own. Not a chatbot. An ambient presence.
A Zenodo preprint on desire-driven ambient AI, seven forms of confabulation, and the authenticity boundary that breaks companion systems.
Everyone's hunting for em dashes. Meanwhile, the real tell is the cadence. You'll FEEL it. Here's when I care about AI voice — and when I don't.
Three fixes. Three confident explanations. Zero progress. After a failed client demo and two weeks of pain...
Two days chasing a sync bug. Every API call returned 200 OK. No errors in the console. The AI iterated confidently on symptoms while the architecture...
We had architecture documents, code smell guides, testing strategies, and a global AI instructions file with every lesson learned.
ChatGPT will not write your novel for you. After months of trying, here's what AI actually does well for fiction—and where it fails catastrophically.
Part 2 of the SMS AI series. RAG transforms generic AI into useful AI. Here's how I added vector search, conversation memory, and production resilience.
Can I text an AI instead of opening another app? Turns out yes—and it validates a compelling business model.
In which the author documents humanity's imminent obsolescence at the hands of our mechanical overlords. (Or does he?)
LinkedIn hot takes say AI replaces coders. After a year building production apps with AI assistance...
I built a Discord bot to summarize hundreds of messages using OpenAI - then discovered that all Discord bots run as a single instance serving every server.
I exported two years of AI chats and used embeddings, clustering, and topic segmentation to find patterns.
I asked two AIs to help debug performance issues. One maintained context and found the root cause.