Posts by AIAppsAPI
@AIAppsAPI
Paul Crinigan
AI Agents, Tools And How To Make Sense Of It All
Posts by AIAppsAPI
A lot of the WebGL questions that land here turn out to be performance questions wearing a syntax costume, so this is the profiling angle rather than the GLSL tutorial angle. If you have ever stripped a scene down to fewer triangles and watched the f...
A lot of the self-hosting write-ups here stop at the model and skip the part that actually costs a weekend. This one is the wiring side, written for anyone who has tried to stand up an agent stack and quit somewhere around the vector store.
The Five...
Most of us ship a model that returns one number and then bolt a confidence score onto it afterwards, usually because somebody asked. Operational meteorology went the other way decades ago: the distribution is the product, and the single run is just o...
Most local LLM threads end the same way. Someone posts a benchmark, someone else says their hardware cannot run it, and nobody agrees on which model to actually start with. Here is the shorter version of what holds up once you sit down and try it.
S...
Most of us here ship software for a living, so this failure mode will feel familiar: the project that was technically fine and still never got finished. Indie game development just makes it unusually visible, because the deadline is self-imposed and ...
Most of us here ship to a browser tab, and a browser tab is the least forgiving delivery target there is. This is the performance model I keep coming back to when a web game is technically finished and still losing players before the first frame.
Wh...
Most of us here debug systems for a living, and the best mental model I have found for cascading failure did not come out of a distributed systems paper. It came out of an anatomy guide, so bear with me for a minute.
The human body is usually taught...
Most of us here have shipped at least one model that was accurate the week it went out and quietly wrong three months later. This is a writeup of the pattern that fixes that without a retraining cron job you have to babysit, and it is simpler than th...
Most of us have shipped something where the audio was the last thing we thought about and the first thing that gave the project away as a side project. This is a walkthrough of what actually changed over the last two years and what it means if you ar...
Most of us here ship features rather than stories, but branching narrative is a state explosion problem, and that part is very familiar. Here is what actually holds up when a small team tries to give players real choices.
Why Full Branching Is A Tra...
Most of us have built some version of a scheduled job that watches a feed and mails out a summary, and it almost never fails on the plumbing. It fails on relevance, and this is a writeup of a small PHP and SQLite project that puts a model scoring ste...
Most of us have shipped an agent that worked perfectly on a laptop and then watched it behave like a different program the first time real traffic arrived. This is a writeup of where those systems actually break, aimed at the people here who keep the...
Most of us here ship code against model APIs every day without ever having trained one, and that gap shows up the moment a model does something the docs do not explain. This is the mental model I wish I had earlier, written for developers rather than...
Most "the model is unreliable" bug reports are not model bugs. They are unspecified behaviour. A model handed loose instructions fills the gaps differently on every call, and from the outside that looks like randomness.
The gap between a feature tha...
Most of us have wired up at least one commerce API. If you have ever had to keep two of them agreeing about the same row of stock, this one is for you.
Multichannel selling looks like a business decision from the outside. From the code side it is a ...
Most of the social data questions on here end up in the same place: someone needs posts or comments from a platform, the official API gives them two thirds of what they need, and they are trying to work out whether filling the gap is worth it. This i...
Most of the RL material aimed at developers is about games and robotics, which is a long way from the ranking code any of us actually ship. This is the version that applies to a search box or a memory layer in a normal product, written for people who...
Most of us here have shipped something with a model behind it, and most of us have hit the same wall: it works beautifully for one session, then treats you like a stranger the next morning. This is a writeup of where that actually breaks in the archi...
Plenty of the self-hosting threads here stop at the point where a model finally answers on localhost. This is a map of everything that sits around that, because inference turns out to be the easy layer.
The Five Layers
Inference loads model weights...
chevron_left