Agents grind until the integration works in the happy path, but the devil is in the details as always.

How does a 5k display replace a 6k display? Are they giving up on 6k? Disappointing.

I feel this should be a bigger focus than it is. All the AI code review start up are mostly doing “hands off” code review. It’s just an agent reviewing everything.

Why not have an agent create a perfect “review plan” for human consumption? Split the review up in parts that can be individually (or independently) reviewed and then fixed by the coding agent. Have a proper ordering in files (GitHub shows files in a commit alphabetically, which is suboptimal), and hide boring details like function implementations that can be easily unit tested.

[0] https://thalo.rejot.dev/blog/plain-text-knowledge-management

[0] https://thalo.rejot.dev/blog/plain-text-knowledge-management

Comments URL: https://news.ycombinator.com/item?id=46807154

Points: 1

# Comments: 0

The format is simple: you define your entities (e.g. opinions, book reviews, facts) including type definitions for metadata fields. Then you write entries that the CLI validates against your schema. The value is the feedback loop: LLMs can easily extract information from unstructured text, but they need validation to make the data high quality. Thalo's "compiler" provides these rules.

Example:

  2026-01-08T14:30Z create book-review "Designing Data-Intensive Applications" ^ddia #book 
    rating: "5" 
    author: "Martin Kleppmann"

    # Summary
    The definitive guide to distributed systems and data architecture. Dense but essential.

It's really up to the user how to use it. Some ways I'm using it:

- End of day journaling: blurt out thoughts, let AI extract insights, facts.

- Organizing my knowledge: using previous "content" I've written (blog posts, websites, my CV) to extract insights.

- Processing commits I've made to keep a living record of my work.

Because it's all plain-text, it allows you to open your knowledge base in any text editor or in Claude Code to do agentic search and analysis. I hooked up a simple Telegram bot to answer questions.

The tooling includes a CLI for validation, an LSP with completions and go-to-definition, a VS Code extension, and a Prettier plugin. There's also a simple scripting API that allows you to loop over entries and use the visitor pattern to traverse your knowledge base or write new rules.

I'd love to hear what use cases people come up with. Let me know what you think!

- Code (MIT): https://github.com/rejot-dev/thalo

- Landing Page: https://thalo.rejot.dev

Comments URL: https://news.ycombinator.com/item?id=46730382

Points: 4

# Comments: 0

An example: in a CST `1 + 0x1 ` might be represented differently than `1 + 1`, but they could be equivalent in the AST. The same could be true for syntax sugar: `let [x,y] = arr;` and `let x = arr[0]; let y = arr[1];` could be the same after AST normalization.

You can see why having just the AST might not be enough for syntax highlighting.

As a side project I've been working on a simple programming language, where I use tree-sitter for the CST, but first normalize it to an AST before I do semantic analysis such as verifying references.

Comments URL: https://news.ycombinator.com/item?id=46706808

Points: 4

# Comments: 0

With all that said, I would still recommend it over anything not retina.

In a project I’m working on I simply present some data and a prompt, the user can then pipe this into a LLM CLI such as Claude Code.

[0] https://modelcontextprotocol.io/specification/2025-06-18/cli...