SteveB retweeted
An outdoor recording, before and after Clear. Sound on. 🔉
After Clear, the background noise in the quiet parts drops from -39dB to -63dB.
Clear runs on device: 5 minutes of audio in 1s on an iPhone 16 Pro, with a 9MB model. It’s in our Swift, Kotlin and JavaScript SDKs.
desertant.com/models/clear
SteveB retweeted
NEW:
- ChatGPT Space for Plus!
- new, simpler visual design across the app
- huge performance improvements everywhere
- completely rebuilt threads– better perf and design
- redesigned Space home
- much faster pages and slides
- edit pages on mobile
- more reliable Codex Remote
much more to come soon!
SteveB retweeted
Introducing @treg_ai /doomscroll skill
My agent watched 272 GTA-6 tips & easter eggs for $0.02 with this
Try it at treg.to/doomscroll
- Name any topic + time window, get the posts worth seeing
- Reads X, TikTok, YouTube, Reddit, HN and Product Hunt at once
- Fully OpenSource
Repo link below 👇
SteveB retweeted
DeepSeek Video AGI moment has come.
OpenMotion, our open-source plugin, now comes with a DeepSeek harness and 300+ video templates inspired by top brands—including Claude, ChatGPT, and Gemini.
@Tuola0205 just used them to make launch videos for your fav five AI products 👇
SteveB retweeted
Cloudflare just dropped Clairic on @huggingface ☁️✏️🧾
Clairic goes beyond decision models: it actually extracts information on a single forward pass, unlocking ultra fast system one decision-making process
▶️ huggingface.co/spaces/huggin…
SteveB retweeted
F*CK it. We're giving away our most-loved speech-to-text model for free.
It's the only pocket-sized model we've seen that can hear your speech well in a NOISY cafe.
It's open-weights on @huggingface today!
- 0.6B params, takes up 600MB space on your device.
- It's TRENDING #2 today among Speech recognition on @huggingface
- SURPASSES State-of-the-art models like ElevenLabs, and DeepGram.
Give it a go yourself, and tell me if it improves your QOL. Either way, I have added the benchmarks below :)
SteveB retweeted
today I'm releasing ORB: the world's least qualified sound designer.
5 MB. completely deaf. never heard a sound in its life. somehow makes a pretty respectable thunderstorm.
try it, it's free:
stdevmac.github.io/orb-audio
SteveB retweeted
5 libraries that kill AI slop (Thank me later)
1.HorizonX
horizonx.so
2. Grainient
grainient.supply
3. Arc UI
uiarc.dev
4. uberdesigner
uberdesigner.ai
5. Bencho
bencho.dev
SteveB retweeted
Researchers have used reinforcement learning to teach a 5-tonne John Deere skid-steer loader to perform complex acrobatic manoeuvres autonomously. Trained in simulation and transferred to a real machine, the system can balance on two wheels, navigate rough terrain, spin, and control its bucket before landing safely.
The research demonstrates the potential of AI-driven automation for heavy construction and industrial machinery.
Video Credit: @leggedrobotics / @ETH
Paper: wevlv.co/4kltYys
#engineering #technology
SteveB retweeted
Meet Allo! 👋
Your friendly neighborhood font, with variable weight and roundness, automatic optical sizing, true obliques, clever and stylish alternates, and loads more thoughtful features and details.
Free & open source: allo.joshpuckett.me
SteveB retweeted
this is the stuff we’re not seeing enough of. we were promised the future but all the labs do from a UI perspective is fancy chatbots. very cool!
SteveB retweeted
are jev demos still cool?
More and more coding agents are going towards the "one manager" thread approach: @cursor_ai projects, @AmpCode spaces, etc.
but the problem with this has always been the classic iMessage problem: you end up having simultaneous unrelated conversations and back-and-forths in the same thread.
So I built a demo using Jev that automatically determines whether it should fork the current chat, or switch to a previous chat, or stay in current one
🧵 some thoughts
Introducing 𝙻𝚒𝚚𝚞𝚒𝚍𝙲𝚘𝚍𝚎𝚌, our ~74M-parameter streaming audio codec powering LFM3-Audio-2B and LFM3-Omni-2B.
An audio codec turns speech into tokens a language model can predict, then converts those tokens back into audio. We built ours around two requirements: tokens that are easy to predict and fast conversion back into speech.
On a single CPU thread on an Apple M4 Max, LiquidCodec runs at approximately:
> 361× real time for encoding
> 87× real time for decoding
Both measurements use FP32 precision.
1/ 🧵
We’re announcing Agent Loop. Self-improving voice agents, now in early access.
Describe the job, set a goal. Agent Loop builds the agent, simulates hundreds of calls, A/B test on live traffic, and fixes what fails until it hits your goal.
Waitlist: vapi.ai/agent-loop
SteveB retweeted
Today we’re launching the Gemini Decisions API, powered by our new Gemini Decision 1 model in @GoogleAIStudio.
Instead of generating text token by token, it’s built for closed-set choices scoring up to 1,000 questions across text, images, audio, video + PDFs in a single pass. A few details:
- 3 decision types: yes/no, multiple choice (up to 255 options) + rubric scoring
- $0.10 / 1M input tokens with $0 billed for output tokens (~5x cheaper than Flash-Lite)
- leads on several benchmark evals with lowest calibration error
Google teams are already building with it, below is a look at Stitch Rapid designing a full product flow in seconds (1x speed)
SteveB retweeted
I'm excited to release Earlybird: a 29M-parameter predictive speech-to-text model that outputs words before you finish saying them.
Most streaming models wait 90-400ms after a word ends before outputting it. Earlybird predicts words correctly a median of 90 ms before they end.
SteveB retweeted
Carbon made a pelican demo so you don't have to 😅
fun test of long-horizon coding and the new /swarm feature in Google @Antigravity.
Builder and critic subagents ran iterative loops across Blender and Three.js - checking their own renders to build a playable 3D game of a pelican biking around our Mountain View campus.
SteveB retweeted
We achieved SOTA on Terminal Bench 2.1 using a DeepSeek v4.1 Flash model.
92.2% at $0.08 cost per task.
Previously Opus 5.5 was the leader at 87.6% at $0.50 cost per task (vals.ai/benchmarks/terminal-…).
Now the obvious question is how did we do it? How can you possibly improve a Flash model to beat frontier models like Opus, Fable, and Astra?
The truth is, we over-obsess about models when they are only half of the equation. The other half is the code that wraps around the model. In fact all of the most revolutionary products (Instinct, Muse, Claude Code, OpenClaw, etc.) are all just harnesses optimizations.
Rather than try to build a new model or fine-tune, we built a harness that would make:
Flash Models —> Frontier Results
Frontier Models —> Beyond Frontier Results
Did I mention it’s also ~6x faster than Claude Code?
And we’re now live. Go Check it out!
Trymidas.io






















