Ask HN: What are you working on? (February 2026)

Two things for my document translator https://kintoun.ai :

1. Trying to improve the translation quality by giving LLM more context.

2. Fixing the issue where PowerPoint slides layout may become a bit messy after transition because of different text density between western and CJK languages.

We're building https://HypeKrew.com/?ref=hn. It is going to be a set of tools for YouTube content creators to better connect with their viewers, based on repeated issues that we've observed when consulting with creators and helping them grow their channels. Right now there's an MVP available, which focuses on

- building an independent line of communication with your audience

- predictive, just in time notifications through push or email delivered when we predict that specific viewer has the time to view videos on YouTube, ensuring you stay on top of their notification stack and don't disappear amongst a flood of notifications.

I'm learning about "AI programming" by working on some toy problems, like an automated subtitle translator tool that can take both the existing English subtitles and a centre-weighted mono audio extracted from the video file and feed it to an AI.

My big takeaway lesson from this is that the APIs are clumsy, the frameworks are very rough, and we're still very much in the territory of having to roll your own bespoke solutions for everything instead of the whole thing "just working". For example:

Large file uploads are very inconsistent between providers. You get fun issues like a completed file upload being unusable because there's an extra "processing" step that you have to poll-wait for. (Surprise!)

The vendors all expose a "list models" API, none of which return a consistent and useful list of metadata.

Automatic context caching isn't.

Multi-modal inputs are still very "early days". Models are terrible at mixed-language input, multiple speakers, and also get confused by background noises, music, and singing.

You can tell an AI to translate the subtitles to language 'X', and it will.. most of the time. If you provide audio, it'll get confused and think that it is being asked to transcribe it! It'll return new English subtitles sometimes.

JSON schemas are a hint, not a constraint with some providers.

Some providers *cough*oogle*cough* don't support all JSON Schema constructs, so you can't safely use their API with arbitrary input types.

If you ask for a whole JSON document back, you'll get timeout errors.

If you stream your results, you have to handle reassembly and parsing yourself, the frameworks don't handle this scenario well yet.

You'd think a JSON list (JSONL) schema would be perfect for this scenario, but they're explicitly not supported by some providers!

Speaking of failures, you also get refusals and other undocumented errors you'll only discover in production. If you're maintaining a history or sliding window of context, you have to carefully maintain snapshots so you can roll back and retry. With most APIs you don't even know if the error was a temporary or permanent condition, of if your retry loop is eating into your budget or not.

Context size management is extra fun now that none of the mainstream models provide their tokenizer to use offline. Sometimes the input will fit into the context, sometimes it won't. You have to back off and retry with various heuristics that are problem-specific.

Ironically, the APIs are so new and undergoing so much churn that the AI models know nothing about them. And anyway, how could they? None of them are properly documented! Google just rewrote everything into the new "GenAI" SDK and OpenAI has a "Responses" API which is different from their "Chat" API... I don't know how. It just is.

I'm thinking all the time about what the "best" way of using local AI agents like Claude / Codex / Gemini is. I'm trying to figure out the best UI/UX. There's so so so much that hasn't been explored yet.

Mainly I'm working on a task dispatch dashboard called Prompter Hawk that is designed to be the best UI for task management with agents. If you've been trying to parallelize by running multiple claude code terminals or codex terminals at once, this tool replaces those terminals and fits them all into one view with an AI task tracking board. It sounds more complicated than it is. It's a harness for Claude / Gemini / GPT models with a GUI that speeds up all your workflows. Rather than using sustained chat mode, all Prompter Hawk tasks are fire-and-forget. You just give the task description and come back when it's done. Parallelism first.

Some example highlight features:

-One dashboard view that shows all your parallel sessions and which tasks each agent has in progress and in their queue. Also shows recently completed tasks and outputs. This is my attempt at the ideal "pilot's cockpit view" for agentic development.

-Tasks are well tracked by the manager: see their status, file changes, and git commits. One click task retry. Get breakdowns on cost per run. Tasks can be set to automatically recur on a given schedule. Everything goes into a persistent local DB so you can easily pull up task data from months ago. Far far better user experience than trying to pull up old chat histories IMO.

-Timeline view and analytics views that give you hard stats on your velocity and how effectively your agents are using and updating your codebase. See unique stats like which of your files your agents read the most and how many daily LOC and commit changes you're doing. See how well you're parallelizing workloads at a simple glance.

-Automatic system diagram generation

-Task suggestion feature. If your agents are idle, they can draft tentative tasks to carry out next, based on the project history and your goals. This makes keeping multiple agents spinning actually much easier than you'd think. You don't need to be a multitasking context-switching god to do this.

I haven't shared it much (not even a Show HN) because the landing page isn't converting well at all yet, though I have some reddit ads doing well. I've had a bunch of free users sign up and a handful of paying users too. Looking for users or just feedback on anything! Sorry for wall of text.

[1] https://prompterhawk.dev/

I'm currently unemployed and I started using Codex a couple of weeks ago so lot's of simultaneous projects, some stalled

Pre-codex:

Local card game: there's a very specific card game played in my country, there's online game rooms, but I want to get something like lichess.org or chess.com scale, oriented towards competitive play, with ELO (instead of social aspects), ideally I would get thousands of users and use it as a portfolio piece while making it open source.

cafetren.com.ar: Screen product for coffee shops near train stations with real time train data.

Post-codex:

SilverLetterai.com: Retook a project for an autonomous sales LLM assistant, building a semi-fake store to showcase the product (I can fulfill orders if they come by dropshipping), but I also have a friend and family order which I should do after this. 2 or 3 years late to the party, but there's probably a lot of work in this space for years to come.

Retook Chess Engine development, got unstuck by letting the agent do the boring busywork, I wish I would have done it without, but I don't have the greatest work ethic, hopefully one day I will manually code it.

Finally, like everyone else, I'm not quite 100% content with the coding agents, so I'm trying to build my own. Yet another coding agent thingy. But tbf this is more for myself than as a product. If it gets released it's as-is do what you want with it.