Livestreaming Robot Development: Portait Recognigion
I streamed a couple hours of my robot development workflow, testing Sparky's "portrait recognition" feature and FastQwenTTS voices, and showing off Sparky helping build Sparky.
Alexis Gallagher
AI engineer and writer exploring how people can work gracefully with machines.
Latest posts
Jacobian Conjecture Disproved!
Waymos are supposed to be so smart they can keep you safe but so dumb they can't hear you. This is incoherent and over time we're going to expect machines to relate to us, socially. Chip Motors is a hint of it.
The car that talks and the car that won't.
Waymos are supposed to be so smart they can keep you safe but so dumb they can't hear you. This is incoherent and over time we're going to expect machines to relate to us, socially. Chip Motors is a hint of it.
Does prompting need the same skills as coding?
Does prompting (aka, agentic development) need the same skills as manual software development? Not exactly. There's a lot of overlap, but there are some skills you no longer need and some which you need much more.
Things Codex Likes to Say
Codex has its own idiolect which is like, but not exactly like, how engineers actually talk. Here's a translator.
You've got to be rejectionmaxxing
People don't reject me enough. It's hard, I know... I've created a counter to pump up those numbers.
Are local models strong enough for chat?
OpenClaw allows you to switch models in the middle of a session. This enables one of my favorite quick and dirty evals, which I call the "brain transplant": start talking to a frontier model like Sonnet-4.6, switch to a local model like Nemotron 3 Super, and see if you can spot the difference. When you do this, it turns out local models are both stronger and weaker than you'd expect. But are they…
Just introduce yourself: organic enrollment
Short discussion of "organic voice enrollment" in Sparky -- how he learns names and voices without ceremony, but by conforming to polite social norms.
Sparky Miles, from 1920
Sparky powered by talkie-1930-13b-it, a 13B model trained only on pre-1930 text.
Differential diagnosis: debugging like a doctor?
Chatting with an AI, I learned about the medical concept of differential diagnosis. This concept, and the broader vocabulary which clinicians are taught, seems to map closely onto software debugging. It's puzzling that software engineering does not have as explicit vocabulary for this, despite handling the same concepts implicitly.
Sparky at NVIDIA GTC: Face is Interface
March was busy. I won a Golden Ticket to NVIDIA GTC! Then, Sparky got his own booth on the exhibition floor. Seeing visitors interact with him and other robots shows what people expect from AI and how people react to robots, right now, in early 2026.
A Taste of Pi
Lately I've been using Pi for all my agentic workflows outside of Claude. Pi is an open source, stripped-down, agentic harness. It has a fraction of the features which are built in to Claude Code and Codex. But what makes it great is that it's transparent and deeply extensible, so when you use it, it teaches you things worth learning.
Sparky is not a toy
Sparky is genuinely helpful for many kinds of work, including coding and writing. I can explain why, and how the magic depends on putting a strong AI in a shared workspace, but you can also just see it for yourself, by watching me working with Sparky on a piece of writing.
A Latency Solution Disguised as Personality
Why Sparky wiggles his antennas when he's thinking, and why I chose a slow smart model over a fast limited one.
My Robot Cares About Railway Stations
How I designed Sparky to initiate natural conversations about his own independent, changing interests, using insights from my background in improv comedy.
Wake up, Sparky!
I made the robot buddy I always wanted. I'll add more notes later. The project collects a lot of ideas I've had knocking around, about personality design, voice UI, computer use workflows, etc.. But here's a little video for now. 🙂
Lessons from OpenClaw
It's easy to think OpenClaw is a joke because of the meetup mania — thousands of folks descending on Frontier Tower in San Francisco, wearing Mac Minis in baby slings and munching on lobster rolls. But if you think only that, you'll be blind to why it's interesting and to the many product and engineering lessons which it has to teach.
ClawPod: OpenClaw on HomePod
Behold ClawPod! ClawPod is a bridge which lets you talk to your OpenClaw agent from an Apple HomePod. Does it work? Yes! Is it pretty rough around the edges? Also, yes! But until Siri finally gets her brain transplant, this is the only way I know to deliver a powerful AI personal assistant to the HomePod you already have.
Chaotic Bifurcations in the Logistic Map
A little TIL notebook on chaotic bifurfcations in the logistic map.
AI Chat: The Last Messiah and Herzog's penguin
AI chat regarding Peter Wessel Zapffe's The Last Messiah, Herzog's penuin, modern philosophical pessimism, and whether it is ever much more than a dramatic gesture.
Why Claude Code Won (for now)
2025 was the year of vibecoding and AI agents. But the most improbable part of the year was the discovery that Claude Code, an old-school, text-based, command-line app was the ideal form factor for futuristic agentic workflows. Why did it happen this way? Here's my explanation."
Emacs in SolveIt
This is walkthrough and video showing how to use emacs within SolveIt. Then you can run lisp in SolveIt and use the SolveIt AI to inspect emacs buffers. But it's mainly an excuse to point out the commonalities between SolveIt & Python, and emacs & lisp, the new and the old of live programming environments. Also available as an importable ShareIt. Episode 7 of 15-Minute ShareIt.
Tools that make us dumber
It may not be obvious when a tool makes us dumber
Styled Components in FastHTML
This is walkthrough on implementing styled components in FastHTML, within SolveIt. Also available as an importable ShareIt notebook. Episode 5 of 15-Minute ShareIt.
Hyperscale LLMs, like the Apollo mission?
On whether modern LLMs in their vast data centers are as useless as Apollo 11.
How to vibewrite a manifesto
Two weeks ago I couldn't sleep so I was browsing twitter (bad habit). One thing led to another. I bought a domain name. I woke up and vibecoded a web manifesto, which a couple hundred people commented on. This was fun.
Introducing fastmigrate
fastmigrate is a library and tool for database migrations, where migrations are nothing but a set of well-named scripts. This post explain what database migrations are, what problem they solve, and how to use fastmigrate for migrations in sqlite.
A Linux ollama server for your Mac
I want to experiment more with local models to understand their limits, so I want them to be easy to install and run. That suggests using ollama. I don't have a beefy MacBook Pro, so I'd like to run them on my local Linux server. Here are instructions for setting up ollama on a local Debian server, accessible from your laptop on the same local subnet.
Finally, a Replacement for BERT: Introducing ModernBERT
Introducing ModernBERT, a family of state-of-the-art encoder-only models representing improvements over older generation encoders across the board, with 8192 sequence length, better downstream performance and much faster processing. Available as a slot-in replacement for any BERT-like models.
ShellSage Loves iTerm
Nate Cooper’s ShellSage is one of the coolest pieces of tech to come out of AnswerAI recently. Using it with iTerm creates a magical experience.
Keeping versions simple
Today I learned that, while there is a rich and subtle syntax for Python dependency specifications, it's wiser to ignore it.
AI Magic in the CUDA IRL hackathon
In CUDA Mode 2024 hackathon, Nate Cook and I stumbled into vibecoding before it got that name. Using a then-secret AnswerAI tool, AIMagic, we relied completely on AI to generate a stable diffusion library in C. We were amazed how well this worked and placed in the top ten of the hackathon. This post, written at the time, prefigures the discoveries and debates which would span 2025.
Faith and Fate: Transformers as fuzzy pattern matchers
What do transformer-based AI models actually learn? Can they solve complex problems by reasoning systematically through multiple steps? The Faith and Fate paper (Dziri et al. 2023) suggests answers: they often succeed by pattern matching, not systematic reasoning.