Jon Iler

Open-source enthusiast. Building 3 businesses with humans + agents. Trying to figure out how to cure my workaholism.

[Jon Iler @joniler]

Spinning up Omapak, an open, community driven Flatpak repo started in Ormarchy-land, but for all distros at its core.

In May 2026 Flathub decided new apps are banned if AI had anything to do with them. Even one AI-touched commit disqualifies a submission. Their house, their rules... but it's a dumb rule, and I'd rather build the alternative than argue with people who are unwilling to have a fair conversation about it.

Normal people don't give a shit how something was built if it works well and it fits their needs. A useless app hand-typed by Linus Torvalds himself is still useless. A great app vibe-coded over a weekend is still a great app. Nobody installs a commit history. If you disagree, I honestly don't care.

Omapak grades the app on its own merits. We don't care as much about how the sausage got made as we do about what the intent is, and what its state is.

- It must build and install, non-negotiable.
- It must not be hostile to the user. (Obfuscated payloads, mystery endpoints, undisclosed telemetry, miner-shaped code: hard pass, pound sand.)
- Packaging isn't a mess. Below 2/5 on packaging hygiene, come back when it installs clean.

Everything else gets scored, published, and never gatekept.

Omapak will be governed in the open, and we're looking for help and submissions actively.

To be clear, I'm not anti-Flathub and I'm not anti-human-review. I'm anti gatekeeping, and anti treating "used an LLM" as a character flaw. That said, Omapak is not a slop repo either. That's the whole point. "Built it using AI/agents" and "slop" are different things.

https://t.co/rOErGvCqqA
@dhh @OmarchyLinux

[Jon Iler @joniler]

@LaNativePatriot This dude is clearly on meth.

[Jon Iler @joniler]

@LaNativePatriot I would have talked him the second he started with that crap

[Jon Iler @joniler]

@W4HO_ Can we get this for Motorola Razr?!

[Brij Pandey @LearnWithBrij]

🚨 A Netflix engineer just open-sourced the fix for the most expensive problem in AI: wasted tokens

It's called Headroom, and it might be the smartest fix to high token usage

Your agent reads a 10,000-token log file to find one error.
You paid for all 10,000 tokens.
The answer needed 1,200.

Headroom sits between your agent and the LLM and compresses everything before the model sees it.

JSON, code, logs, RAG chunks, each gets its own specialized compressor.

And it's reversible: the originals stay on your machine, so nothing is lost.

The results speak for themselves:

→ Up to 95% fewer tokens
→ Same accuracy on benchmarks
→ Zero changes to your code

Setup takes one minute:

1. pip install "headroom-ai[all]"
2. headroom wrap claude

Done.

Works with Claude Code, Cursor, Codex, and anything OpenAI-compatible.

Everything runs locally, fully open source.

The cheapest token is the one you never send.

🔗 Github repo: https://t.co/SSMujMiwMy

So, are you compressing your context, or just paying the bill?

[Surajit Dasgupta @surajitdasgupta]

That Gen Z is the first generation that is dumber than its predecessors is no longer an opinion. It is scientifically established, as is the cause.

Listen to neuroscientist, educator and bestselling author who specialises in human learning and brain development, Dr Jared Cooney Horvath:

[Jon Iler @joniler]

@ElitzaVasileva Dynamic workflows will max you out fast.

[Jon Iler @joniler]

@webenstein_ I like Fedora a lot.

[Mia @MiaAI_lab]

Rejoice single DGX Spark owners!

You can now run Qwen3.8 Flash Next on TensorFold with the new Nvidia FP4 QAD quant 🔥

- 1.1M total FP8 KV cache
- 256k context, 5 concurrent users
- 46-55 tok/s on prose, single stream
- 132 tok/s on prose, 5 concurrent steams
- Prefill 2000-2300 tok/s

Qwen3.8 Flash Next is still the BEST model to run on one DGX Spark, bare none!

You can choose what to run: FP4 QAD for quality, or int4 for speed. Details on what is QAD below.

Get it here:
https://t.co/ElQiyWOMdQ

[Zach Siegel @zacherysiegel]

AI can build the app.

It can't tell you it feels wrong.

Taste is the moat.

[Jon Iler @joniler]

@xPraveen07 Supabase is just Postgres with extra steps.

[Tenzen Studio @tenzenstudio]

Excited to launch a fully featured Illustrator alternative, but 10x better.

Another huge step in creating the new, no-subscription creative cloud. While others focus on cloning the entire suite simultaneously, I decided to go one by one, but go deep each time.

MCP, skills and CLI are available so you can create awesome designs using your agents.

[Jon Iler @joniler]

Here's a big idea, Adobe: try differentiating yourself by not being massive pieces of shit! Then maybe people wouldn't have to reverse engineer your products because they're tired of dealing with your garbage company.

[Jon Iler @joniler]

@LoopandPixels Waste management

[Jon Iler @joniler]

@omarvvvr This guy clearly doesn't have a dev server box lol. Herdr my dude.

[Zach Siegel @zacherysiegel]

AI didn't make details matter less. It made them the whole game.

Anyone can generate an app now. The ones worth using are made by people who polish the last 10% and still keep the big picture in their head.

[Mia @MiaAI_lab]

GLM 5.3 FLASH is now FREE in OpenCode 🤯

This adds to the already free DeepSeek v4.1 Flash.

[Zach Siegel @zacherysiegel]

Every agent app is turning into the same screen: a list of sessions with a status dot next to each one.

That's an inbox.

The next big agent UX win won't be a better chat box. It'll be better triage.

[Diogo Chaves @diogochaves]

There was no qBittorrent app for Omarchy that looked and felt like a real app. And since I run qBittorrent in Docker, double-clicking a .torrent or clicking a magnet link did nothing, unless I installed the desktop qBittorrent too.

So I made one: omaseed.

[Jon Iler @joniler]

@haydendevs You're buying freedom and owned infrastructure, once you go all in you'll never want to go back. Get Usenet.

[Jon Iler @joniler]

@faqdev 0%