1,128 karma · joined October 31, 2012
I'm currently working on dev tools for agentic engineering and software-first simulated hardware design. Writing and researching as a means of growing.
Projects:
- Intraview (intraview.ai) give Agents the ability to co-create code tours in VS Code to speed up onboarding and alignment on code-bases
- Unfudged (unfudged.io) is source control without the commits so Agents (and you) can recover from mistakes.
- Writing / research (cyrusradfar.com) is where I do market research, guesstimating the future and keep searching the glorious unknown canvas of knowledge.
Live in Portland, OR. Previously lived/worked in both DC, and SF.Find all my projects, writing and contact methods at https://cyrusradfar.com
I’m an Iranian American developer. The US is bombing my parents’ country. Before the bombs, sanctions already cut Iranian developers off. Gilens and Page found US citizens have “near-zero” impact on their own government's policy decisions. If Americans can’t control their government, why punish 90 million Iranians for theirs? Sanctions are the Versailles playbook: they destroy the middle class, the only group that ever drove reform from within. Call 844-228-6422 and tell Congress to vote on the War Powers Resolution.
This is off-topic, but I'm curious on your interest in Zig and experience getting started. I haven't explored the language, yet. Currently, I'm focused on absorbing the rust ecosystem a bit better.
Given the scale of code moving, I'll need to do a full manual regression so I can build, going forward, off the new setup.
My goal for that release is this Friday and if I miss that It'll be the next Wednesday.
Thank you.
One install, one init, and then it just works. It shouldn't stop across restarts or crashes.
It doesn't read the file the instant the OS fires the event. It accumulates events and waits for 3 seconds of silence before reading. So if an editor does write-tmp → rename (atomic save), or a tool writes in chunks, we only read after the dust settles.
I accept there are cases if the editor crashes mid-state that you have a corrupted state but there was never a good state to save, so, arguably you'd just restore to what's on file and remove the corrupt partial write.
It's not bulletproof against a program that holds a file open and writes to it continuously for more than 3 seconds, but in practice that doesn't happen with text files by Agent tools or IDEs.
Feel free to follow up for clarity.
I think it would be dishonest if I didn't share that your approach to discourse here isn't a productive way of asking what insights I'm bringing.
If that's your concern, I agree I can't claim that nothing exists to solve pieces of the puzzle in different ways. I did my research and was happy that I could get a domain that explained the struggle -- namely unfucked.ai/unfudged.io -- moreover I do feel there are many pieces and nuances to the experience which give pause to folks who create versioning tools.
I'm open to engaging if you have a question or comment that doesn't diminish my motives, assumes I must operate in your world view "problems can only be solved once", and discourages people to try new things and learn.
Look, I'm grateful that you stopped by and hope you'll recognize I'm doing my best to manage my own sadness that my children have to exist in a world where folks think this is how we should address strangers.
I could build an extension for the UI vs a Tauri app, and it could help you install the CLI if you don't have it. Would that meet your needs?
That said, the fidelity of OS-level daemon can't really be replicated from within an app process.
We posted show the same day to solve the similar problems.
My solution https://news.ycombinator.com/item?id=47172238 (unfucked.ai) works with any agent and any file in the folder whether it's edited by the agent or anyone by tracking all file writes.
Amazing how this problem was top of mind for all of us at the same time!
UNF is one install command + unf watch to protect a repo on every file change, takes 30s.
Time Machine snapshots hourly, not on every change, so you can lose real work between snapshots. This may have changed or I missed something but I reviewed that app to see if it was possible.
And while tmutil exists, it wasn’t designed to be invoked mid-workflow by an agent. UNF* captures every write and is built to be part of the recovery loop
Would make my eve to hear any folks who are going to cancel in protest of this continued consolidation.
To frame this post, my intent isn't to be yet another post on Anthropic and OpenAI calling out Chinese labs.
I was interested in the pattern, which isn't new. Edison, Apple, the record labels, and even Disney did it.
Build on everyone else’s work, then lobby to stop anyone from building on yours.
Was hopeful for a discussion on more history and, potentially, more solutions that could help us be more innovative and less exploitative.
Chime in, would love more perspectives.
I didn't open up the source for this as I have a mono-repo with several experiments (and websites).
Happy to open the source up and link it from the existing website.
I've started to have an Agent migrate it out, and will review it before calling it done. Watch https://github.com/cyrusradfar/homebrew-unf
Edit: You can download the current version now: https://github.com/cyrusradfar/homebrew-unf/archive/refs/tag...
> unf watch
# reboot
> unf list
it should say watching on your directory still, if it stays crashed or something else. ping me at support at v1.coJust one human, two machines at my home can't replicate all configurations...
If an AI agent rewrites 30 files and you haven't touched jj yet, jj has the before-state but none of the intermediate states. UNF* captured every save as it happened, at filesystem level.
jj is a VCS. UNF is a safety net that sits below your VCS.
- UNF* works alongside git, jj, or no VCS at all
- No workflow change. You don't adopt a new tool, it just runs in the background
- Works on files outside any repo (configs, scratch dirs, notes) as it doesn't require git.
They're complementary, not competing.W.r.t. to the histogram, this is my fav feature of the app as well. Session segmentation (still definitely not perfect) creates selectable regions to make it easier, too. The algo is in the CLI as well for the Agent recap (rebuilding context) features.
From what I know (correct me) magit-wip-mode hooks into editor saves. UNF hooks into the filesystem.
magit-wip-mode is great if your only risk is your own edits in Emacs. UNF* exists because that's no longer the only risk; agents are rewriting codebases/docs and they don't use Emacs.
I am more interested in testing if folks have the problem and like the shape of the solution, before I try to decide on the model to sustain it. Open Source to me is saying -- "hey do you all want to help me build this?"
I'm not even at the point of knowing if it should exist, so why start asking people to help without that validation.
I work(ed) with OSS projects that have terrible times sustaining themselves and don't default to it bc of that trauma.
Thanks for stopping by.
I spent a bit of time being baffled nothing existed that does this. Then I realized that, until Agents, the velocity of changes wasn't as quick and errors were rare(er)
The fact that there are countless use cases for real government efficiency to help the people they would sacrifice because Anthropic wanted to refuse killer robots is baffling.
1. Can a company change their ToS to clarify that you can't "hack" their Oauth for other tools -- Yes, no question
2. Should they enforce the ToS, absolutely
The elephant in the room is Google has no support infrastructure and setup no notice drip campaign before they started banning out-right.
They could've easily rate-limited or even blocked requests authed from apps outside Anti-gravity, but banning entire accounts is pretty ridiculous. It'll just lose them very eager AI enthusiasts who are the audience they NEED to grow their new experiments.
It's bad business and not in their best long-term interests. Big fail. This is the way a startup may act because they're going broke on traffic -- not Google.