Urgent.News

What's breaking now, across thousands of outlets.

AI

I let my own 31B model take over development of the thing running it

Two things I build: Flash , a local coding agent that runs in your terminal. Flash Onyx , the model it runs on. Gemma4 with a system prompt and sampling baked in. As of today, I develop both of them from inside Flash, with Onyx driving. /model natuworkguy/flash-onyx-2.3:31b-cloudbase That tag is 96 KB on disk, because there are no weights in it. It is my prompt and my sampling pointed at a hosted…

In an unprecedented development, I entrusted my own 31B model, Flash Onyx, to take control of its own development process. The model, which runs in a terminal-based Flash agent with a system prompt and sampling baked in, operates on the Gemma4 architecture. The entire setup, weighing in at a mere 96 KB, relies solely on the system prompt and sampling to a hosted base, eliminating the need for hefty weights on the machine. The 12B variant, if desired, can run fully locally.

The core of the system is a system prompt, Onyx's obedience to which is documented in a modelfile within the repo. If the model's actions prove troublesome, I issue corrective instructions directly within the same session. This approach, coupled with the Python-based tools integrated within the model's tree, allows for seamless operation and iteration.

Unlike traditional evaluation suites that provide numerical outputs, the true value of this setup lies in its ability to expose subtle, systemic issues. These are not bugs, but subtle tics that can be detected only by human observation over prolonged periods. The most significant rule to emerge from this experiment is a self-policing mechanism that prompts the model to verify its actions before proceeding.

With a built-in safeguard against overstepping its bounds, the model is designed to request confirmation before committing to any changes. This ensures that the model remains accountable and does not quietly write itself a permission slip. The model's development branch, 2.4, incorporates these safeguards, with a new rule mandating evidence for every 'DONE' line in a report and demanding that shell calls be explicitly stated.

The project, accessible on Ollama and hosted on GitHub under an MIT license, is now live and ready for others to run and experiment with. With the model operating under a self-imposed, iterative governance model, I am keen to hear about any potential pitfalls or unexpected behaviors that others may encounter.

Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at dev.to →

More in AI

When Should an AI Suggestion Become a Financial Record?

AI can classify an expense, propose a match between a bank transaction and an invoice, or flag an unusual cash-flow pattern. None of those outputs should automatically become accounting truth.

  • Store AI suggestions as immutable artifacts with key information
  • Commit suggestions through domain rules, not just confidence score
  • Maintain separate layers: suggestion, decision, and record layers

Agentic AI Is Mostly Marketing. Memory Is the Part That's Real.

Open any 'agentic AI' repo trending on GitHub this month and you'll find the same shape: a while loop, a tool-calling wrapper, a retry policy, and a system prompt that says 'you are an autonomous…

  • Most "agentic AI" in 2026 is automation with chat interface, not truly agentic.
  • Memory allows systems to change behavior over time without human intervention.
  • Many projects add memory as an afterthought, resulting in plateaued performance.

From API to GPU, Week 6 (Part 2): Watching a Neural Network Learn

Phase 2 of 8: Enough ML to understand inference. Week 6 of 32, part 2 of 2. Part 1 built a one-neuron model that converts Celsius to Fahrenheit, made it predict, and scored how wrong it was with a…

  • Neural network learning involves weight and bias adjustments
  • Loss is minimized through gradient computation
  • Optimizer updates occur during training loop stages

The Dependency the Agent Invented: A Myth-Busting FAQ

Consider a reconstructed incident that shows up when teams treat an agentic loop as an architecture review. A backend pair watched their coding agent close six steps with a confident summary and a…

  • A successful loop only shows code consistency, not external dependencies
  • Multi-step traces indicate scenery familiarity, not dissenting opinions
  • Zero exit code means process completion, not external object mutation

Mistral Small 3.2 Lands With Sharper Function Calling and a 128K Context Window

Mistral just pushed Mistral Small 3.2 (24B) to Hugging Face, and the timing is interesting. While OpenAI keeps nudging GPT-4o mini's pricing and Anthropic gates Claude Haiku behind API quotas, Mistral…

  • Mistral Small 3.2 model introduced with 24 billion parameters and 128K context window.
  • Enhanced function calling with v3 tool schema format, 78% accuracy improvement.
  • Open-weight model allows fine-tuning and deployment without additional fees.

I Love AI. I Still Don’t Want It in Everything.

I love AI. I use it constantly, I experiment with new models, I build things with it, and I genuinely think there are a lot of places where it can make software better.

  • Author loves AI but dislikes its forced inclusion in every app
  • AI should enhance product only if it genuinely improves functionality
  • Optional AI disabling preferred to avoid unnecessary features

More from Saturday 5 September →