RSSAmplifier

Blog

Little Short Bulletins

Likely Some Balderdash

/RSS feed ↗10 posts

Latest posts

Welcome In, now showing in the Epistemologies Of Slop show in San Francisco

Welcome In is a greeting in the age of free slop, a programmatic greeting with a performance component. It just went up at Tiat and is on display through the end of July 2026.

An LLM that talks in acrostics

Sidechat is a exploration of sidechannels in LLM chatbots. The chatbot can answer questions, as usual, and can also have another conversation, out in the open, under the radar.

Visualizing tiny LLMs, in parameter golf

How does a 16MB LLM behave?

Driving coding agents for hours, using test-driven dev

Coding agents do more when they can guide themselves. Building an app that an agent can drive and debug, and building a test suite that an agent can use for directions, allows agents to continue to take turns steering themselves to the correct answer.

DSPy on a Pi: Cheap Prompt Optimization with GEPA and Qwen3

Case study: chat-to-SQL

Running Stable Diffusion in under 400MB in the browser in under half a second

Stable Diffusion XS, with a ControlNet based on sketch, can generate 2-4 images per second on a consumer grade desktop in the browser, at a total size (text encoder, controlnet, u-net, autoencoder, all the pieces together) of under 400MB when compressed.

Diffusion Local Time, extra-large and extra-small, at CVPR

Load Diffusion Local Time XL in a dedicated window → Diffusion Local Time, extra large Diffusion Local Time is an art piece that displays the time in surrealistic landscapes. At CVPR, I displayed two versions, Diffusion Local Time XL and Diffusion Local Time XS. Extra large is precomputed on a desktop computer with tweening from minute to minute like the Dali Clock! 09:44 on Diffusion Local Time…

Diffusion Local Time, a timepiece made of generative AI art, at Art Hack Day

Diffusion Local Time is a functional timepiece that explores surrealism in generative AI. It was on display in the exhibition at Art Hack Day in San Francisco. In this installation, the underlying technology and the politics remained the same (a Latent Consistency Model ControlNet running locally) and the local computer was upgraded to a high-end desktop computer, for a sub-minute render latency.…

Diffusion Local Time: Stable Diffusion running on a Raspberry Pi as an AI Art Timepiece

What is this Diffusion Local Time is a timepiece with a Generative AI display. A Raspberry Pi locally generates and displays pareidolic clock faces every several minutes, using open-source code and open-source typography and freely-available models . Easy to customize The clock faces are generated from four text prompts that default to California landscapes, and this is easily field-serviceable to…

Stable Diffusion Clock

Three clock faces! Open source! Check it out!