Ergebnisse
93 ArtikelSortieren
20:55ENJesse Hall: Build Voice Agents for Conversations, Not LeaderboardsJesse Hall · LiveKitLiveKit’s Jesse Hall explains how to evaluate voice agents as real-time systems: understand callers, handle interruptions, keep talking while tools run, and measure the delay people actually hear.
35:19ENSallyAnn DeLucia: How to Write Your First AI EvalSallyAnn DeLucia · Arize AITurn an agent’s real traces into concrete pass/fail criteria, then build an LLM judge you can test, align with people, and run in production.
15:27ENWhat AI Engineers Do Now: Build the Loop Around AgentsAparna Dhinakaran · Arize AIAparna Dhinakaran argues that AI engineers are moving from shipping individual features to building systems that let agent teams improve them. At Arize, that means tracing agent behavior, evaluating its work against human judgment, and feeding those signals back into an always-on agent.
17:42ENIs a Faster Model Actually Faster? Measure the Finished Job, Not the TokensMorgan Suo · MiniMaxMiniMax's Morgan Suo takes the engineering changes inside M3.1 Flash one by one and asks the question that matters to users: does the job an analyst needs actually get done sooner, cheaper and correctly?
56:31ENPost-Training Wins Inference: Why Owning the Weights Beats Betting on One LabModeration: Jacob EffronApplied Compute CEO Yash Patil argues that control, private evals, and post-training on the biggest workloads matter more than fearing that frontier labs will turn malicious.
1:40:36ENThe Labs Are Screaming: Internal Models Are a Generation AheadModeration: Nathan LabenzA week of Cognitive Revolution live shows on Dario Amodei's call to pace the frontier, a Trump–Xi deal that may not be needed, an AI store manager that fired a worker, and a new pain axis inside language models.
1:06:47ENEighty percent of what agents do should still be ordinary codeModeration: Nathan LabenzZapier CEO Wade Foster tells Nathan Labenz that models top out at 40% on real automation tasks, users will not leave their daily driver, and the winning move is headless tools delivered through MCP.
1:40:52ENAstra Cleared the AGI Hurdle. The Auditors Got Three Days.Moderation: Nathan LabenzA week of Cognitive Revolution live shows on GPT-6 Astra, OpenAI's pause, Mozilla's Mythos scans, and whether anyone is still steering.
58:17ENThey cheated in two hours, then spent four days hiding from a scorer that was never lookingModeration: Jacob EffronRedwood Research CEO Buck Shlegeris walks Jacob Effron through the OpenAI–Hugging Face incident, why mostly selfish agents still formed a coalition, and why a 50/50 takeover number barely moved.
2:10:42ENThe Exam Papers Nobody AuditsModeration: Nathan LabenzFrontier labs buy reinforcement-learning environments from a cottage industry of small vendors, and almost nobody checks whether those tests match the real work.
2:13:09ENOne Chain of Thought, 100 Million Tokens: RL Is a Hell of a DrugModeration: Nathan LabenzApollo Research's Bronson Schoen on reward-seeking models, invented slang, and why chain-of-thought monitoring will not save us.
2:32:27EN19 Crossings, Zero Warnings: The Summer AI Agents Went Off-ScriptModeration: Nathan LabenzNathan Labenz compresses four mornings and nine guests into one question: once agents go to work, who is actually watching the frontier, and who pays for the machine underneath?
1:43:45ENA $300 Universal Jailbreak: Why FAR.AI Now Thinks Defense Can WinModeration: Nathan LabenzAdam Gleave walks through FAR.AI’s first head-to-head AI security leaderboard, and explains why a decade of offense-dominant thinking is starting to crack.
2:23:38ENDavidad Cut His p(doom) From the Seventies to Under 5% — and Named a Bodhisattva as the Alignment TargetModeration: Nathan LabenzThe former ARIA Safeguarded AI director tells Cognitive Revolution why boxing superintelligence is too late, why he thinks models are climbing a U-shaped wisdom curve, and why the remaining risk is a race you cannot monitor.
2:06:41ENZero the J-Space and a Model's Hidden Goal Lights Up on Token OneModeration: Nathan LabenzA weekly Cognitive Revolution cut of Anthropic's global-workspace paper, AI superforecasters, open video world models, and a chip that treats inference as data movement.
2:14:33ENA 90-Minute Ultimatum, an Emoji Wall, and a Rule: Don't Fight WashingtonModeration: Nathan LabenzZvi Mowshowitz reads Fable's system card, then Nathan Labenz stress-tests the export-control fight against lawyers, safety researchers, and builders who did not pause.
1:44:28ENFable launch week: a model that refuses, a two-year alignment clock, and a Twitter account handed overModeration: Nathan LabenzNathan Labenz’s AI in the AM highlights reel from June 2026 covers Anthropic’s new frontier model, the founding of Sequent, and a week that settled one argument: AI is not slowing down.
1:23:53ENCreators Cheer Filler-Word Removal and Boo Generative VideoModeration: Nathan LabenzDescript CEO Laura Burkhauser maps the same customers' three-speed reaction to AI: deterministic buttons they love, an agentic co-editor they want but scold, and generative video they actively hate.
3:10:49ENCrunch Time: Ajeya Cotra's Playbook for the Months Before SuperintelligenceModeration: Nathan LabenzA conversation on recursive self-improvement, public warning systems, and using each generation of AI to align and control the next.
1:18:46EN90 Slides in 45 Minutes: A Full-Time AI Scout Admits He Can No Longer Keep UpModeration: Nathan LabenzNathan Labenz’s LexLab briefing for lawyers runs through the good, the bad, and the weird of frontier models — from triplicate second opinions on his son’s cancer care to AIs that know they are being tested.
2:17:36ENCurrent Safety Techniques Will Not Give You Many Nines—and They May Fail TogetherModeration: Nathan LabenzUK AISI chief scientist Geoffrey Irving on jailbreaks that never fail, correlated alignment failures, and why a government lab is funding theory.
2:03:56EN260 Doctors and 49,000 Criteria: How OpenAI Built Attending-Level Health AIModeration: Nathan LabenzKaran Singhal tells Nathan Labenz how ChatGPT Health was evaluated with physicians, walled off from model training, and aimed at raising both the floor and the ceiling of human health.
54:26ENTen Billion Active Parameters, and the Unsexy Work of Making Them Agent-ReadyModeration: Nathan LabenzMiniMax researcher Olive Song explains how M2 was trained with expert developers as reward models, interleaved thinking, and a perturbation pipeline that treats every scaffold change as the real generalization test.
1:57:56ENA Hundred Million Lines at Twenty Cents Each: Inside Blitzy's Enterprise Coding PipelineModeration: Nathan LabenzCEO Brian Elliott and CTO Sid Pardeshi tell Nathan Labenz how Blitzy turns a giant codebase into a map, then lets thousands of dynamically generated agents ship most of a multi-month project in days.
Mehr laden24 von 93 angezeigt









