Ayooluwa

here to learn and share
building @aethexailabs

[Josh Meyer @_josh_meyer_]

Introducing VAmoS Bench from @veris_ai 🔥

3,300 calls, 11 agents, 100 scenarios

We compared what agents said with what they did

@pipecat_ai @trydaily @livekit @Vapi_AI @elevenlabsio @cartesia @OpenAI @GoogleAI @retellai @NVIDIAAI @DeepgramAI

https://t.co/lf3iFhB8DO

🧵

[AethexAI @aethexailabs]

Introducing the AethexAI React + React Native SDK.

Build cross-platform conversational AI agents for web, iOS, and Android in minutes. WebRTC and Expo support from day one.

Blog: https://t.co/LLJwHLBEi8
Github: https://t.co/r60YfwgqoN
Docs: https://t.co/A65dzYLvSG

[Ab. @Abiodun0x]

Some news: Café One, the largest co-working space chain in Nigeria, will begin rolling out Decide today as an exclusive member benefit across its nationwide network.

This is the first major rollout under Decide for Work, our deployment arm for bringing spreadsheet agents into companies, co-working spaces, and universities.

[AethexAI @aethexailabs]

Oubliez la démo : déployez gratuitement votre propre agent vocal de haute qualité en cinq minutes. Conçu pour les marchés émergents et pour une échelle mondiale, stack complète incluse.

[AethexAI @aethexailabs]

Skip the demo, ship your own high-quality voice agent for free in five minutes. Built for emerging markets and for global scale, full stack included.

[AethexAI @aethexailabs]

One platform, scoped API keys, signed webhooks, and a runtime built for the narrow tolerance of a live call. https://t.co/ps0vVwVZ0j

[TechCrunch @TechCrunch]

These two founders left Goldman and Meta to build voice AI for markets everyone else overlooked https://t.co/a8dowJ3gSt

[AethexAI @aethexailabs]

Yesterday, AethexAI came out of stealth and introduced Kora 1, our voice AI stack for enterprise deployment across emerging markets.

Today, we’re sharing a closer look at our enterprise platform and the voices powering it.

[AethexAI @aethexailabs]

These two founders left Goldman and Meta to build voice AI for markets everyone else overlooked https://t.co/LdmJDNsCld via @techcrunch

[MWC @MWCHub]

Meet some of the experts & leaders who will be joining us at #MWC26 Kigali! 🇷🇼

We can't wait to welcome speakers from key companies, such as @GSMA, @AXIANTELECOM, @ericsson, @Vodacom, @aethexailabs, @AminiTech, @RwandaSpace + more.

Explore the agenda 👉https://t.co/KGgJ20uIAg

[Lukas Ziegler @lukas_m_ziegler]

🚨 BREAKING: Genesis AI has just launched GENE-26.5, what the company is calling the first robotic brain to give robots human-level physical manipulation capabilities.

The demo video alone is extraordinary, robots cooking 20-step meals, solving Rubik's Cubes mid-air, playing piano, and conducting lab experiments with delicate instrumentation. All with human-level dexterity.

@gs_ai_ has built a dexterous robotic hand that exactly mirrors the human hand, paired with a data collection glove with tactile-sensing electronic skin.

When a human wears the glove, every movement maps 1:1 to the robotic hand.

This closes the embodiment gap. In this case human skill transfers DIRECTLY to a robot at scale.

The economics are game-changing too, the glove is 100x cheaper than typical options and delivers 5x greater data collection efficiency vs traditional teleoperation. That makes continuous large-scale robotics training viable for the first time.

The company has raised $ 105M in seed funding backed by Eclipse, Khosla Ventures and Bpifrance, with Eric Schmidt and Xavier Niel among the strategic angels.

And the first general-purpose robot is coming soon.

Co-founded by @zhou_xian_ and @theo_gervet, this is a full-stack robotics company controlling every layer, AI, hardware, simulation, and data.

That's a serious moat.

~~
♻️ Join the weekly robotics newsletter, and never miss any news → https://t.co/GoA3ZuwoPB

[Claude @claudeai]

With the Autodesk Fusion connector, designers and engineers can create and modify 3D models through conversation.

[a16z @a16z]

For all of history, humanity shared one 3D world.

@theworldlabs co-founder @drfeifei says spatial intelligence now lets us generate and reconstruct infinite universes for creativity, travel, storytelling, even socialization.

“It suddenly will enable us to live in a multiverse way.”

[Anish Moonka @anishmoonka]

Went down the rabbit hole on this. There are bacteria in your gut right now with tiny electric motors built into them. Each motor is 45 nanometers wide, about 2,000 times thinner than a human hair. It spins faster than a Formula 1 engine. After 50 years, scientists just cracked how it works.

The motor spins a corkscrew-shaped tail so the bacterium can swim. At that tiny scale, water feels as thick as tar. Moving anywhere takes serious power. A single E. coli cell (the kind in your gut) spins its motor at 18,000 RPM. That beats modern Formula 1 engines, which redline around 15,000. Some bacteria in the ocean run theirs at 42,000 RPM, nearly triple.

And the motor barely wastes any energy as heat. Your car engine loses most of its fuel to heat. This thing loses almost none.

Inside the motor, 5 proteins form a ring wrapped around 2 proteins in the middle. Five can't split evenly into 2. The resulting lopsidedness is what makes the whole thing work. Protons, which are tiny charged particles, get pulled from outside the cell through the motor. Each one grabs a center protein, then lets go. In letting go, it tugs the outer ring a fraction of a turn. Another proton does the same thing on the other side. Then another. It's like two feet alternating on bicycle pedals. Over 2,000 times per second.

Switching directions is a whole other trick. When the bacterium senses food running out, it tags a small messenger protein with a phosphorus atom. That tagged messenger floats over and touches one protein on the outer ring. The touched protein flips into a new shape. That flip triggers the next protein, and the next, and the next, around the whole ring, like dominos falling. The ring reshapes in milliseconds. Rotation reverses. The bacterium turns and swims somewhere else.

Mike Manson, a biophysicist at Texas A&M, has been studying this one motor since the 1970s. For five decades, most of its parts stayed a mystery. Starting in 2020, a new wave of imaging let scientists see the individual pieces. The last pieces clicked into place in a March 2026 paper from Aravinthan Samuel's lab at Harvard. Manson told Quanta Magazine his lifelong quest was fulfilled.

A billion years of evolution built the most efficient rotary motor on the planet. Trillions of them are spinning inside you right now.

[Natalie Wolchover @nattyover]

Bacteria move around using a molecular machine called the flagellar motor that rotates faster than the flywheel of a race car engine and switches directions in an instant. After 50 yrs, scientists have finally figured out how it works. “My lifelong quest is now fulfilled.” Link⤵️

[Elon Musk @elonmusk]

🇺🇸🇺🇸

[Aksel @akseljoonas]

Introducing ml-intern, the agent that just automated the post-training team @huggingface

It's an open-source implementation of the real research loop that our ML researchers do every day. You give it a prompt, it researches papers, goes through citations, implements ideas in GPU sandboxes, iterates and builds deeply research-backed models for any use case. All built on the Hugging Face ecosystem.

It can pull off crazy things:

We made it train the best model for scientific reasoning. It went through citations from the official benchmark paper. Found OpenScience and NemoTron-CrossThink, added 7 difficulty-filtered dataset variants from ARC/SciQ/MMLU, and ran 12 SFT runs on Qwen3-1.7B. This pushed the score 10% → 32% on GPQA in under 10h. Claude Code's best: 22.99%.

In healthcare settings it inspected available datasets, concluded they were too low quality, and wrote a script to generate 1100 synthetic data points from scratch for emergencies, hedging, multilingual etc. Then upsampled 50x for training. Beat Codex on HealthBench by 60%.

For competitive mathematics, it wrote a full GRPO script, launched training with A100 GPUs on https://t.co/udm7xGpNzR, watched rewards claim and then collapse, and ran ablations until it succeeded. All fully backed by papers, autonomously.

How it works?

ml-intern makes full use of the HF ecosystem:
- finds papers on arxiv and https://t.co/brvCC7fLPa, reads them fully, walks citation graphs, pulls datasets referenced in methodology sections and on https://t.co/hrJuRkRyzi
- browses the Hub, reads recent docs, inspects datasets and reformats them before training so it doesn't waste GPU hours on bad data
- launches training jobs on HF Jobs if no local GPUs are available, monitors runs, reads its own eval outputs, diagnoses failures, retrains

ml-intern deeply embodies how researchers work and think. It knows how data should look like and what good models feel like.

Releasing it today as a CLI and a web app you can use from your phone/desktop.
CLI: https://t.co/l3K1PslZ1n
Web + mobile: https://t.co/orko5srL4H

And the best part? We also provisioned 1k$ GPU resources and Anthropic credits for the quickest among you to use.

[vixhaℓ @TheVixhal]

Being able to finally understand and do the math you saw when you were a little kid is one of the best feelings in the world.

[vixhaℓ @TheVixhal]

https://t.co/UOX9HBErXe

[Aadi Kulshrestha @MankyDankyBanky]

I trained a 12M parameter LLM on my own ML framework using a Rust backend and CUDA kernels for flash attention, AdamW, and more.

Wrote the full transformer architecture, and BPE tokenizer from scratch.

The framework features:
- Custom CUDA kernels (Flash Attention, fused LayerNorm, fused GELU) for 3x increased throughput
- Automatic WebGPU fallback for non-NVIDIA devices
- TypeScript API with Rust compute backend
- One npm install to get started, prebuilt binaries for every platform

Try out the model for yourself: https://t.co/TB2itlmCVT

Built with @_reesechong. Check out the repos and blog if you want to learn more.

Shoutout to @modal for the compute credits allowing me to train on 2 A100 GPUs without going broke

cc @sundeep @GavinSherry