David Stout
563 posts
David Stout
@Davidstout
Founder, AI Scientist, CEO of webAI™
Austin, TX เข้าร่วม Mart 2020
2.8K กำลังติดตาม11.1K ผู้ติดตาม
Two of the top three RAG models, courtesy of @thewebAI
David Stout@Davidstout
@thewebAI now has 2 of the top 3 RAG models on ViDoRe. And both can run on an iPhone or your laptop. Our smallest model (4B) scores within 0.1 nDCG of @nvidia's nemotron-colembed-vl-8b-v2. That’s within margin of error. Same retrieval quality. Half the model. For a long time, multimodal RAG looked like a scaling problem: Bigger models More GPUs More data center spend That assumption just broke. We’re now seeing: - frontier-level retrieval (64.48 - 63.49 nDCG) - real multimodal understanding (documents, not just text) - running locally This isn’t just a leaderboard result. It means: - your data doesn’t need to leave your device - latency drops to device speed - marginal cost goes to zero - retrieval becomes private by default The best RAG models now fit in your pocket. And we’re making them available. So anyone can build, run, and own their retrieval stack locally. Available: lnkd.in/ejxvZyFe lnkd.in/ejfgVBYP
English
@thewebAI now has 2 of the top 3 RAG models on ViDoRe.
And both can run on an iPhone or your laptop.
Our smallest model (4B) scores within 0.1 nDCG of @nvidia's nemotron-colembed-vl-8b-v2.
That’s within margin of error.
Same retrieval quality.
Half the model.
For a long time, multimodal RAG looked like a scaling problem:
Bigger models
More GPUs
More data center spend
That assumption just broke.
We’re now seeing:
- frontier-level retrieval (64.48 - 63.49 nDCG)
- real multimodal understanding (documents, not just text)
- running locally
This isn’t just a leaderboard result.
It means:
- your data doesn’t need to leave your device
- latency drops to device speed
- marginal cost goes to zero
- retrieval becomes private by default
The best RAG models now fit in your pocket.
And we’re making them available.
So anyone can build, run, and own their retrieval stack locally.
Available:
lnkd.in/ejxvZyFe
lnkd.in/ejfgVBYP
English
More open source announcements this afternoon with @thewebAI .
This one is very special.
#ai #opensource
English
Excited to announce that webAI is partnering with the @USArmy on Project Aria, contributing to Team Black. AI for soldiers, on the edge, on the devices they already carry. Let's go!
David Stout@Davidstout
Proud to partner with the @USArmy on Project ARIA. @thewebAI is contributing to Team Black — building a "Model Armory" that delivers custom AI to soldiers at the tactical edge, on the devices they already carry, no cloud required. Sovereign AI isn't a talking point. It's a mission requirement. webai.com/blog/webai-sel…
English
Proud to partner with the @USArmy on Project ARIA.
@thewebAI is contributing to Team Black — building a "Model Armory" that delivers custom AI to soldiers at the tactical edge, on the devices they already carry, no cloud required.
Sovereign AI isn't a talking point. It's a mission requirement.
webai.com/blog/webai-sel…
English
English
I can't speak for them, but to speculate with the best understanding of the situation.
*First, I saw reporting that they were extremely close on a deal, and for much of the time both sides really wanted to reach one. I have seen what happens in tense negotiations when things get stressed and deteriorate super fast, and I could believe that was a large part of what happened here.
*We believe in a layered approach to safety--building a safety stack, deploying FDEs and having our safety and alignment researcher involved, deploying via cloud, working directly with the DoW. Anthropic seemed more focused on specific prohibitions in the contract, rather than citing applicable laws, which we felt comfortable with. We feel that it it's very important to build safe system, and although documents are also important, I'd clearly rather rely on technical safeguards if I only had to pick one.
*We and the DoW got comfortable with the contractual language, but I can understand other people would have a different opinion here.
*I think Anthropic may have wanted more operational control than we did.
English
I'd like to answer questions about our work with the DoW and our thinking over the past few days. Please AMA.
English
David Stout รีทวีตแล้ว
David Stout รีทวีตแล้ว
Narrative Violation: “Job Postings For Software Engineers Are Rapidly Rising”
English
I love my country. The people defending it shouldn't have to negotiate with a tech company for access to AI.
@AnthropicAI is the surveillance state it's warning you about.
I wrote a response.
webai.com/blog/ai-archit…
English
David Stout@Davidstout
"Stack Mac Minis at home" is just the mainframe argument wearing a turtleneck. There was a moment for it — maybe two years ago, when the only models worth running were too large for a single device to load. That moment passed. Machines got better. Models got smaller. @thewebAI was the first to do this, but we stopped ... AI models are shrinking. Intelligence compresses. I'm running the world's best coding model on a single M2 Mac right now — no cluster, no rack, no ritual. Models will be commodities. Edge AI will be table stakes. And no real consumer or business is going to babysit a tower of hardware in their kitchen. The future isn't more boxes — it's AI that disappears into the device you already own. Private. Safe. Effortless.
QME
I spent $72,000 on 120 Mac Minis, and another $12,000 setting everything up for OpenClaw.
I want to change the world, and I believe this ClawdBot setup can do that.
Will keep you guys posted on how everything develops.
So far I've used my agents to reply to three emails and filter another 268 that were unread, which is a 10x productivity improvement from what I typically accomplish each day.
English
"Stack Mac Minis at home" is just the mainframe argument wearing a turtleneck.
There was a moment for it — maybe two years ago, when the only models worth running were too large for a single device to load. That moment passed. Machines got better. Models got smaller. @thewebAI was the first to do this, but we stopped ...
AI models are shrinking. Intelligence compresses. I'm running the world's best coding model on a single M2 Mac right now — no cluster, no rack, no ritual.
Models will be commodities. Edge AI will be table stakes. And no real consumer or business is going to babysit a tower of hardware in their kitchen.
The future isn't more boxes — it's AI that disappears into the device you already own.
Private. Safe. Effortless.
English
English
Yes @jaminball! Systems of Record are Dead. Long Live Systems of Record! 👑💀 Agents will happily automate total chaos if your source of truth is fuzzy. Turns out rock-solid data rules & boring governance just became enterprise AI’s sexiest superpower. 🚀 cloudedjudgement.substack.com/p/clouded-judg…
English
LLMs are the new disk drives: commodity infrastructure you hot-swap for whoever’s cheapest + best. The fantasy that the model is a moat just expired. ☠️
English
Modern generative AI is just executing a kinematic equation.
It treats intelligence as a ballistic trajectory. The model activates all neurons equally, or worse, uses a central router (MoE) to assign work.
This is Central Planning.
No central router can effectively manage the complexity of general intelligence at scale.
We are building the alternative: The Agoric Transformer.
English
This is so awesome.
Michael Dell 🇺🇸@MichaelDell
$6.25 billion. 25 million children. $250 each. Susan and I believe the smartest investment we can make is in children. That’s why we’re so excited to contribute $6.25 billion from our charitable funds to help 25 million children start building a strong financial foundation through Invest America. 💪📈🇺🇸 onedell.com/investamerica/
English
The next wave of AI isn’t about bigger models in the cloud — it’s about smaller, faster, more specialized intelligence running close to the user and the data.
English