AI has a reliability problem. We're fixing it. 💪

Thanks @Nasdaq for the shoutout today 🥳 Read more about our $100M Series A: lnkd.in/gVbjmR3N

Hallucinations are AI's biggest unsolved problem. General-purpose models have to boil the ocean — handle everything from quantum physics to dinner recipes. That's exactly why they can't be trusted with anything critical. We built our leading model, APT, differently. 👇

"These frontier models are amazing—they're sort of like schizophrenic geniuses. They can create incredible answers, and then you can ask the same question a second time and get a completely different answer that might not even be correct." — @roth_dan, CEO of Scaled Cognition. Thanks to @Steve_Rosenbush and @WSJ for sharing our story. Read the Exclusive: wsj.com/tech/ai/scaled-cogni…

What happens when you optimize your AI agent for customer satisfaction? Say a shipping company deploys an LLM trained to get thumbs up. Someone calls asking where their lost package is. The system can admit it's lost or say it's coming tomorrow. Saying the latter would make the customer happy and the agent would earn a thumbs up for lying. @profdanklein on Gradient Dissent: that's not a bug, but a reward function working exactly as intended. Full conversation in the comments.

Our CEO @roth_dan joined @vkhosla on @FoxBusiness today where @LizClaman asked what got @vkhosla so excited about @ScaledCognition. He summed it up perfectly: "It was pretty simple, reliability is as important as intelligence when it comes to AI systems" 🎦 Full interview: lnkd.in/g7ApKRd9 🅰️ scaledcognition.com/series-a

I’m excited to announce @ScaledCognition’s $100M Series A led by @vkhosla and @KhoslaVentures! I’ve spent decades in NLP, and the gap between fluent and true is exactly where AI breaks down. @roth_dan and I put together an incredible team to build a verifiable architecture focused on truth, controllability, and alignment. The result: models that are smaller, faster, less expensive, and much more reliable.

Certain applications just can’t afford hallucinations. Scaled Cognition is so far the best answer I have seen. Best customer support solution I have seen.

LLMs didn't just introduce new errors — they erased the signals that told us something was wrong. Fluency used to mean something. Now a model can be maximally fluent and completely wrong. On why the smell is gone, and what that means for AI reliability: scaledcognition.com/blog/the…

The best AI models can be brilliant one moment and completely wrong the next. That's not an intelligence problem, it's a reliability problem.

The promise of autonomous AI agents: they independently execute workflows, make decisions, and act without hand-holding. The problem no one's talking about: an agent that's mostly right can't be trusted to act alone. Our Co-Founder & CTO @profdanklein is speaking on exactly this tomorrow at @awscloud Summit Los Angeles. "The Next Frontier: Enterprise AI Agents That Work Alone" 📍 Startup Zone, Booth #112 🕐 12:30 PM PST | June 10th aws.amazon.com/events/summit…

Most teams measure hallucinations by catching the obvious — a wrong number, a made-up product. But a well-written policy from the wrong customer tier? A summary with the right shape and the wrong details Those don't get caught. Those get shipped.