Top
Best
New

Posted by david927 23 hours ago

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

What are you working on? What have you been curious about lately?
290 points | 1009 commentspage 17
clickclickhero 2 hours ago|
I am building a small utility tool to add events to calendar from text/speech/image https://text2event.com/
nbpname 6 hours ago||
Privacy & Security with self-hosted LLMs! You have dreamed of this too?

Today you you can self-host some models, yet this does not mean you can safely browse the web with it, and automate tasks that would interact with your email account, your bank account and your clients.

This is the reason why I left Mozilla, to work full-time on fixing these problems.

Prompt-injection, hallucinations and private information leakage might not be fixable, but they can be identified 100% of the time, deterministically, before any undesirable consequences.

https://cogilibre.com/

Godsend69 3 hours ago|
To identify potential issues with your self-hosted LLMs, consider using The Shield. It allows you to see what your LLMs are sending out and seal your AI models, ensuring a safer browsing experience. https://lightinthedarksolutions.com/shield.html
nbpname 11 minutes ago||
Thanks for caring about this topic as well.

However, asking people to pay first without an established brand nor the ability to audit the code requires a lot of trust. Without any additional judgement about the product, I would never recommend anybody to do that by fear of spyware, phishing or worse …

From the same consideration, I intend to open source my solutions. At the very least offer auditing capabilities to people like me, who are too paranoiac about security and privacy.

On the topic, I do not want to ask the user to always vouch for every usage. My goal is to detect ahead of executing any command, and offer options to re-try automatically, strip the leaking context, abstract the information or accept it. Only asking the user when there is a doubt, while not bothering the user when there is no issue.

matheusmoreira 13 hours ago||
I'm doing a massive AI code review on my lone lisp project. I finished a complete Fable code review, and the experience pissed me off enough I switched from Anthropic to OpenAI. Decided to run an equivalent code review with Sol and compare it with Fable. I'll publish the results on my website when it's done. Some pretty interesting results already. I think I stressed Anthropic's safety classifiers to the max.

Meanwhile, I decided to revive my Rust liblinux project. Rust has come a long way since 2019, notably the inline assembly macro has been stabilized. I will learn Rust this time!

wbnns 2 hours ago||
I left my job at the end of June; I'm tired of playing someone else's music and want to write my own songs (build my own apps)

So I'm building an app a week and focusing on solving niche problems I've experienced in my own life to start

I think it's never been easier to build prototypes, but because of all of the slop, building community and finding users around them are harder than ever before

This latter part I think is going to be the hardest part of going out on my own

But going to try

Things I'm working on right now:

https://a2ciple.pt/ - testing prep for a language exam required for residency & citizenship in Portugal

https://basehub.org/ - a docs site for agents to get the latest context on Base, an Ethereum L2

https://playgokab.com/ - a vocab game to learn words in diff languages

https://mru.space/ - a protocol to run software that gracefully degrades across failing hardware

https://shiponeshot.com/ - a framework to help me ship prototypes more quickly

Anyhow, cheers :)

narayanahari 7 hours ago||
We’re building Deckhand (https://getdeckhand.dev), an AI assisted, human in the loop platform for shipping and operating mobile apps.

We have built a few apps recently and what we observe is that while building the app has gotten dramatically faster with AI, everything around it still takes a surprising amount of human effort - testing, debugging, reviewing, store submissions, releases, monitoring, and keeping things running.

Deckhand is our attempt to make that entire loop faster, while keeping humans in control. Would love to hear how others are shipping and operating mobile apps.

g58892881 21 hours ago||
Photo upscaling and inpainting apps, both running in the browser https://crisp.photos https://wipe.photos
forsalebypwner 20 hours ago||
Sweet, been looking for a better browser-based inpainting tool, will check it out.
jjcm 20 hours ago||
Simple and efficient. Nicely done.
g58892881 11 hours ago||
Thank you
hackitup7 14 hours ago||
I was only working on this occasionally, but I built an extension to block sites in a more creative way.

Basically I'm finding myself in a bind where:

- I spend too much time doomscrolling (X, reddit, HN etc) - AI is making me lose touch with a lot of basic skills like coding, and I was already losing touch with doing real math

So I built an extension that blocks sites and forces you to actually solve a problem before you can get to a distracting site. The idea is to 1) block the reflex of visiting a site and getting a dopamine hit, and 2) at least get some mild cognitive benefit from it. I use it myself.

https://sphinx.fyi/

idontneedcoffee 19 hours ago||
I am building my "dream tool" too, a cross-device roaming profile-experience context-focused UI. Many, many iterations that go back maybe 15y, even built a custom linux distro (iolinux) back in the day(ro root, containerized user-spac with composer-like app handling). Always worked on side-projects, consulting gigs or one-off implementations, often times across multiple customer-devices. Switching context is damn expensive, hence tried to build a tool to help with that. Always failed on the UI parts - native linux tooling was just awful. What do you specifically mean by "context" you ask? Well, you get a call, switch to work on `work://customer-foo/project-bar/task-baz`, every bound device / tool / application loads data related to that task (your filesystem mount shows relevant files, browser loads relevant tabs, obsidian relevant notes etc). Building my own small db engine to accommodate that use-case, getting more fun working on it every day
dv35z 14 hours ago||
Hey - spent a lot of time in this world (solution / sales engineer, technical account manager to many clients), and thinking about this same problem.

Even experimented with per-client Linux users. Per-client calendars & routines. Per client knowledge-base, etc

Very curious to see what you come up with.

idontneedcoffee 9 hours ago||
I divide my unstructured universe into workspaces, each workspace is a self-contained shareable/moveable soon-to-be-containerized runtime running an in-process LMDB with roaring bitmaps and (for now, S2-like hilbert space projections will replace it soon) lancedb for some basic cossim search. You configure your data sources per Workspace - imap, files, s3, soon slack(bc I use it), on top you create a classically-looking virtual tree. That tree is made of bitmaps - /work/customers/acme org/projects does a logical AND on all bitmaps in the path (work AND customers AND..) - if you zoom out to /work/customers you see all data for all customers, if you'd create /work/projects - you'd see all data tagged with any project. On top of that tree sit contexts - think moveable db views. You create a context "acmeorg" and bind your browsers, cli(for dynamic dotfiles), notes apps etc to it. You can even mount a file representation of all indexed objects via a fuse driver. When you point acmeorg at /work/customers/acme org/projects/foo everything moves with you - your browser stashes current tabs and opens relevant ones, your fs mount shows only relevant content etc - moving to projects/bar does the same, if you do that via commandline ctx set ../projects/foo --update-dotfiles - you get your dotfiles You can setup actions - if I send a browser tab thats a yt url into home/music download it and place it into my roaming profile workspace Home, if I index a arxiv website, hit the library agent to properly assign it to existing tree paths and write a summary There is also timeline support - you can have a personal and wikipedia aaand historian foo timeline, wanna know the zeitgeist of when your grandma was born - tick extra timelines and query in layered mode Ou, and there is real-time streaming support, your db can resurface documents based on GPS or based on your camera feed (or, will implement today - show related documents based on your desktop view) - primitive vector sim search now, will be something much better riding the internal representations of gemma4 (for now) - ok let me stop here, you'll have a pinned task list per workspace on the left and a set of a2ui canvases and old-school boring and rigid apps that will either be directly contextualized or just ride the contextualized filesystem(as obsidian does). Ou, did I mention agent usecase? Agent switches to /work/task1 and his local *.md files are only related to that task, want more zoom, switch to /work. ok, oook, let me really stop here
mapontosevenths 17 hours ago|||
Reminds me a bit of qubes OS with different motivation.
calvinmorrison 16 hours ago||
This is awesome. I always have imagined my computer being well, my desk. So far I have to live with a bunch of i3 screens I name after whatever I'm working on. I'd love to be able to page through the bench to flip to a task. Unfortunately with apps going so mono-windowed (slack, etc) it's harder to do. If I could just say have my slack window for this client, with their docs, proposals, database, whatever up, that'd be sick. I love the idea thank you. good luckl
idontneedcoffee 9 hours ago||
replied to another comment - if you are curious go over it, thx!
gghootch 21 hours ago||
https://nommer.ai turns recipes into 2-player mode so you can cook together

People seem to love the concept so far and are starting to sign up, now just need to figure out how to get them to actually use it

kofta999 20 hours ago|
Currently building Tensou in my free time

https://github.com/kofta999/tensou

I always get frustrated when using local file transfer tools like:

Dukto: was my personal favorite, but is tiring to run cross platform, and most importantly does not support transfer resumes (so if copy got cancelled/disconnected at 99% you either have to copy again or search for the diff and copy it manually)

LocalSend: pretty good cross platform support, but does not work when connecting 2 devices using a LAN cable, and resume support is poorish

So I'm building Tensou to be an alternative that doesn't compromise all these features. Supporting resumes, folder syncing similar to rsync, cross platform (well I don't have a Mac to test on but it should work, and mobile support is planned)

Performance wise it's on par with Dukto and even slightly better with large files (and I'm still working on optimizations for small files too)

Would like to hear HN's feedback or what other features should I add? My goal with Tensou is to be the best, most performant and feature rich local file transfer tool ever.

More comments...