Left panel is the ecology where evolution happens. Right panel is a maze. I pick an evolved organism and drop one cell into the maze. Some genomes fail. Some explore the whole thing. Zero control after injection.

The cells don't have functions like 'move left' or 'eat food'. Each cell runs a small evolved gene network that reads local inputs and writes to registers. Physical consequences follow from the register values. The cell doesn't know it's navigating. Its internal chemistry just happens to produce movement.

~2000 lines of C. Single thread. Runs on a laptop

Comments URL: https://news.ycombinator.com/item?id=47457782

Points: 2

# Comments: 0

I mitigate it by giving the agent a fixed action set (no scripts, no direct API calls), and breaking tasks into focused subtasks so no single agent has broad scope. The LLM prioritises its own instructions over page content, but if someone managed to hijack it, the agent can interact with authenticated sessions. Everything's visible in real time though, and all actions are logged, so you can see exactly what it's doing and kill it.

Practically speaking, I use it similar to how people use Zapier or n8n, you set up specific workflows and make sure you're only pointing it at sites you trust. If you're sending it to random unknown websites then yeah, there's more risk.

But even then, an attacker would need to know what apps you're authenticated with and what data the agent has access to. The chances of something actually happening are pretty low, but the risk is there. No one's fully solved this yet.

So the model isn’t carrying the whole instruction chain across multiple steps, it’s just solving the current task. Similar pattern to what tools like Codex CLI or Claude Code do.

Just need to see if people find this version useful

In this demo I gave it my CV and asked it to find matching jobs. It scans my inbox, opens the listings, extracts the details and builds a Google Sheet automatically.

Comments URL: https://news.ycombinator.com/item?id=47322046

Points: 4

# Comments: 9

I wasn't looking for these. They appeared in my feed. I signed up as a normal user and opened dev tools

Comments URL: https://news.ycombinator.com/item?id=47083597

Points: 3

# Comments: 3

Comments URL: https://news.ycombinator.com/item?id=47040033

Points: 1

# Comments: 3

It isn’t a transformer or LLM — it doesn’t generate code. It extracts the structural pattern between two snippets and compiles a deterministic rewrite rule. Same input → same output, every time.

Examples: • console.log(x) → logger.info(x) generalises to console.log(anything) • require(“x”) → import x from “x” • ReactDOM.render → createRoot • custom project conventions

The rules apply across an entire codebase or through an MCP plugin inside Claude Code, Cursor, or plain CLI.

It runs entirely on CPU and learns rules in real time.

Tool: https://hyperrecode.com I’d really appreciate feedback on the approach, design, or failure cases.

Comments URL: https://news.ycombinator.com/item?id=46202949

Points: 2

# Comments: 0

Show it console.log(x) -> logger.info(x) and it learns the pattern, then applies it across your entire codebase. Deterministic, same input, same output, every time.

Not a transformer, not generative, not probabilistic. It parses code into AST, extracts the structural pattern, and executes exact rewrites.

Works as an MCP plugin for Claude Code, Cursor, and Claude Desktop.

https://hyperrecode.com

Would love feedback.

Comments URL: https://news.ycombinator.com/item?id=46199665

Points: 3

# Comments: 0

You’ve got a frozen transformer and a second module still trained with SGD, so how exactly does that solve forgetting instead of just relocating it?

It is not an LLM and it is not a template system. There is a small learned component that decides when a rule applies, and the rest is a deterministic structural rewrite engine.

There are a few demo modes:

TEACH: learn a structural rule from examples COMPOSE: apply several learnt rules together TRANSFER: use the same rule in different symbolic domains SIMPLIFY: multi step rewriting CODEMOD: for example you can teach lodash.get to optional chaining from two examples

Once a rule is learnt it generalises to inputs that are deeper or shaped differently from the examples you gave. Everything runs on CPU and learning happens in real time.

Demo: https://re.heavyweather.io

Comments URL: https://news.ycombinator.com/item?id=46174215

Points: 2

# Comments: 1