Handler is a Mac app for Codex and OpenCode that adds a review layer while the agent is generating code.

Every edit comes with a short explanation: what changed, why it changed, and what it touches. If one edit needs more context, you can open a side chat on just that change and ask follow-ups without derailing the main agent run.

The goal is to build a mental model as the code is generated, instead of waiting until the end and reviewing one giant diff cold.

It’s the same idea that makes stacked PRs easier to review: smaller, understandable units. Handler applies that at AI generation time, before the change turns into one giant diff.

This year I mostly stopped writing code by hand. I drive agents now, and I’m faster for it.

But a few months in I noticed a problem. The agent would come back with a 400-line change, tests green, diff looked reasonable, and after the first few chunks I’d pretty much lose the mental model. At that point I wasn’t really reviewing anymore. I was skimming the shape of the change and moving on.

I know that’s not the ideal way to drive an agent. But I also don’t want to slow the whole thing down and manually inspect every line like it’s 2022. The whole point of using these tools is speed.

So the bottleneck stopped being writing code. It became understanding the code fast enough to stay in control of my own project.

Handler is what I built to fix that for myself:

- Every edit the agent proposes comes with a plain explanation of what that specific change does and why. - The explanation is attached to the edit, so you build context incrementally instead of reading one huge summary at the end. - You can pull any single edit into a side chat and ask “why this approach?”, “what else does this affect?”, or “redo this part (soon)” without blowing up the main agent context. - You go edit by edit and accept or reject, instead of staring at one wall of diff and hoping. - It runs on your existing Codex and OpenCode setups, with worktrees so parallel runs don’t collide. - It also has a built-in terminal that feels closer to a Chrome DevTools console than a plain terminal: command output is easier to inspect, the agent can read the terminal state, and there’s a native JSON viewer/builder for working with structured data without copy-pasting into another tool.

What it’s not: a review bot like CodeRabbit or Greptile. I use those, and they catch real bugs. But they review the code after the fact.

Handler points earlier in the workflow. It helps you review while the agent is producing the change, so you can keep the mental model instead of trying to reconstruct it later from a finished PR.

I’ve been using it daily and can’t go back to merging agent changes blind. If you’ve felt the same problem with large AI-generated diffs, I’d love for you to try it and tell me where it falls over.

Comments URL: https://news.ycombinator.com/item?id=48367682

Points: 4

# Comments: 0

I know a product like this will get banned on HN, but I think the technical part is interesting, so I’m taking my shot here.

problem: You try to get ChatGPT to write your landing page copy or email and it comes back... cringe. It's generic. It's as if every output is written by the same person. You try adjusting the prompt and running it through the AI again, and it's just a different version of the same boring output.

There's no real creative exploration going on here. It's one model, one shot, one voice.

solution: Rather than relying on a single AI to compose your content, I created a system with over 100 different AI personas, each with their own area of expertise, personality, and aesthetic (based on real-world data), to rate and score your content in a variety of ways. And then, took some inspiration from AlphaEvolve (Google DeepMind's evolutionary coding agent) [1], we take these personas as a fitness function and apply an evolutionary algorithm to your content in a variety of ways. It’s a search problem, not a one-shot problem.

The result: Copy that's been stress-tested by a diverse panel and evolved through selection pressure. Not just whatever one model generated on the first try.

Would love to hear your feedback.

1:https://deepmind.google/blog/alphaevolve-a-gemini-powered-co...

Comments URL: https://news.ycombinator.com/item?id=47432531

Points: 3

# Comments: 0

I built a PDF reader that can generate interactive visualizations right inside the document. The goal is to reduce the “intuition gap” when reading dense papers: select a concept, click Visualize, and it tries to produce a solid visual interactive app where you can rotate/zoom/step through (not just a text explanation).

Upload any PDF, select text, click Visualize. Demo (no signup): https://zerodistract.com/try/pdf/67cdee74-810b-4f1b-af7d-010...

Product link: https://zerodistract.com

I’d love feedback, especially on what feels useful vs. distracting, and where it breaks.

Comments URL: https://news.ycombinator.com/item?id=47183256

Points: 8

# Comments: 2

Comments URL: https://news.ycombinator.com/item?id=43934562

Points: 6

# Comments: 2

https://news.ycombinator.com/item?id=35826540

https://news.ycombinator.com/item?id=35510949

Comments URL: https://news.ycombinator.com/item?id=42790955

Points: 4

# Comments: 2

https://news.ycombinator.com/item?id=35826540

https://news.ycombinator.com/item?id=35510949

I'm excited to introduce Graphthem, a search engine designed to explore the deeper layers of knowledge rather than just surface-level popularity.

While many AI search engines simply summarize top N results, we've found this approach often misses the many good stuff that is buried deeper in the links and references.

Graphthem takes a different approach. we don't just look at the first few pages we find. We also dig into what those pages link to, so you get the whole story. This allows us to deliver answers that capture not just what's immediately visible, but also the foundational ideas and deeper insights that inform those results.

Here's a comparison between Perplexity and Graphthem:

Query 1: "most watched youtube videos?"

Perplexity: https://www.perplexity.ai/search/most-watched-youtube-videos...

Graphthem: https://graphthem.com/search?uuid=7f0839a9-85ee-4f81-a5c9-cb...

-

Query 2: "what is pagerank?"

Perplexity: https://www.perplexity.ai/search/what-is-pagerank-ieJ.4iFqTq...

Graphthem: https://graphthem.com/search?uuid=c689e9d1-d3d4-4d51-8602-00...

---

Some related ideas I've explored before:

https://news.ycombinator.com/item?id=35826540

https://news.ycombinator.com/item?id=35510949

I would love to hear your feedback. Please let us know how we can improve.

Comments URL: https://news.ycombinator.com/item?id=42540979

Points: 13

# Comments: 5

Comments URL: https://news.ycombinator.com/item?id=42214104

Points: 5

# Comments: 0

I am temporarily pausing and making sure I am within legal limits. If not, I will completely remove face recognition and try other routes to solve the problem.

Vignesh

Edit:

Fixed!

Exactly - this won't be useful for stalkers. We don't crawl or index social media pictures.

Will add FAQ section.

Clearview vs Introthem:

- Clearview does photo-to-photo matching. We don't do that, and I don't think I will ever build that.

- You have to provide the name, then we build the faces collection for analyzing at search time and delete it.

- We don't retain any face collection once the search is done.

I still don't know if I am breaking any laws, but here is how Introthem works.