Enterprise AI strategy, platforms and multi-agent research
Building intelligent, private systems
Building enterprise AI platforms and researching multi-agent systems. Most of the work comes back to the same question: what can run privately, and what do you end up renting from somebody else.
Focus Areas: Enterprise AI Platforms · LLMs & GenAI · MLOps · AI Governance · Multi-Agent Systems
I write at jeremysnr.github.io about AI infrastructure, sovereign compute and the industrial decisions Britain is making, or failing to make.
Latest: Your Kids Will Either Build the Next Economy or Rent It From America
📦 snugZero-dependency TypeScript primitive for fitting prioritised content into a token budget. Core of the snug ecosystem. |
Phase 1 · Text-level blackboard hive with cross-inhibition and a structurally protected dissenter (Carol). Agents debate, challenge each other, and converge — with one voice that cannot be silenced. Phase 2 · Agents communicate through a shared continuous-valued latent buffer rather than text, using BAPC codecs, TIES-Resolve conflict resolution, and damped fixed-point iteration. Emergent collective representations confirmed at GPT-2 and 7B scale. |
|
Ecosystem of adapters for the snug token budget library — drop-in fit for OpenAI, Anthropic, and tiktoken. |
🔀 convergeZero-dependency TypeScript primitive for converting LLM message arrays between provider formats. OpenAI ↔ Anthropic ↔ Gemini. Pure data transformation, no API calls. |
|
AI color grading for Adobe Lightroom Classic. Describe a look in plain language, hand it a reference photo, or let it choose for you. Ashla reads your image and applies the result as a develop preset. |
