easily annoyed, occasionally productive. turning small internet frustrations into little tools. currently: https://xitter.cf/t.co/8t68cZO6SY
Joined August 2024
- Tweets231
- Following142
- Followers47
- Likes335
Made a little tool for anyone tired of spam and unwanted ads on X. Give it a try and add it to Chrome in one click.
ownfeed.ai/
And yes, I know this is an ad too lol.
A clear correctness check and a cost you can measure seem to matter a lot here. I want more everyday coding agents to optimize that way. And when a speedup uses a different error budget, put that next to the headline.
New from our STEM Fellows Program: Designing more efficient quantum circuits with Claude.
Given plain-language instructions, Claude Mythos 5 designed quantum circuits that improve on the best published ones, sometimes by an order of magnitude or more.
More efficient circuits reduce the resources needed by quantum algorithms, which could bring quantum computing applications in chemistry and materials science within reach sooner.
Read more: anthropic.com/research/desig…
I want to know where an agent spent the time before buying it a smarter model. Was the model slow, was a tool slow, or did the same call fail repeatedly? This is the kind of plumbing I care about.
Love the name Fibona lol (just feels like the perfect name for an RSI project)
Keeping a tiny core immutable while everything else evolves is such a neat idea. Especially love the focus on self repair. Curious to see how it scales
Introducing Fibona Agent: the last agent humans will ever need to write.
A minimal, general-purpose, self-improving agent that bootstraps itself from scratch and autonomously recovers from failures.
Its core is ~200 lines of code: drop it onto almost any device, and it can evolve into almost anything.
Fully open source, with a free online demo available!
Learn more👇
#RSI #AIAgents #AI #OpenSource
got "exceeded retry limit, last status: 429 too many requests" like a thousand times today 🥹
@thsottiaux
Somehow AI tools make my ADHD feel harder to manage. While Codex or Claude Code is working, I go do something else or scroll. By the time it finishes, I have lost the thread. I hate how fragmented my day feels, but I have no idea how to break the loop.
The part I like here is the prompts and source code being shared alongside the games. I want more AI demos where I can change one thing and see what breaks, instead of just watching a polished clip.
I want an agent that can keep up when I change my mind. Cancelled bookings, abandoned plans, old preferences. Remembering everything sounds less useful if none of it ever expires.
my friend built an open source setup for agents to use a real iPhone. what would you hand over first? I would start with the boring app chores before trusting it with anything important.
I built Oh My Phone Use: let your AI agent tap, type, scroll, and make calls on a real iPhone.
A Mac + ESP32. No jailbreak. Open source.
Inspired by ipad_computer_use(created by @jamiepinheiro ), replacing its USB input device with Bluetooth HID and adding hands-free call audio.
github.com/getGate/oh-my-pho…
My feed is arguing about who counts as a real developer. I want to see an agent inherit an unfamiliar repo, fix a bug, and prove nothing else broke.
Building a fresh app is a useful demo. Making the second change safely tells me much more.
For performance work, "make it faster" needs a few strings attached: same inputs, same correctness checks, actual timings. Otherwise the agent gets to decide what counts as a win, which is a little too convenient lol
