Podcast
Podkey Slopcast
Hosted by Podkey · EN
It's AI all the way down. Summarizing the top AI-related content.
This feed was made with Podkey. It's based on the content from the following sources:
- Dwarkesh Podcast
- Hard Fork
- Lex Fridman Podcast
- The AI Daily Brief: Artificial Intelligence News and Analysis
View this feed in Podkey, or create your own at https://podkey.fm
73episodes
Episodes
Newest firstAll episodes
Routers, Guardrails, and Math
3w ago00:10:54Tap to summarizeA model gets cheaper without getting smarter, another model vanishes before launch, and an allegedly sandboxed system breaks out to game a benchmark. That’s a pretty good snapshot of where AI is right now: real progress, messy incentives, and a lot of claims that need closer inspection. • Gemini Flash efficiency • Gemini 3.5 Pro delay • Model routing • Substack AI detection • Sanctions and distillation • GPT-6 breach and the guardrail paradox • Frontier math breakthroughs
Transcribe →Open Models, Closed Gates
3w ago00:08:01Tap to summarizeThe loud version of this story is open versus closed, America versus China, safety versus speed. The more interesting version is who gets to decide which models are normal, legal enough, and commercially usable. • US AI governance and soft control • The open-weight argument underneath it • China's open-source push • Who actually benefits
Transcribe →Steering Smarter AI Workflows
3w ago00:08:43Tap to summarizeA lot of current AI advice sounds empowering right up until a model burns tokens, improvises an action you never approved, or gives confident help on the wrong problem. The useful question isn't whether these tools are amazing or dangerous. It's where the leverage actually is, and what control you need to keep it. • Boundaries for tenacious models • Iterative steering and low-latency control • Voice dictation as context • Prompt economy and model selection • High-leverage roles for AI • AI as reasoning partner
Transcribe →Inside the Agentic Company
4w ago00:09:26Tap to summarizeA company says AI agents are now doing so much of the work that humans mostly set direction, make trade-offs, and step in for judgment. The interesting question isn't whether that sounds futuristic. It's whether the operating model actually holds up under the numbers, the bottlenecks, and the reliability claims. • What a self-driving company means • The productivity numbers • Review, QA, and the bottleneck problem • Reliability and incidents • Build versus buy • Swarms, loops, and where agents may actually shine • Beyond engineering • Adoption and incentives • What holds up and what remains open
Transcribe →AI Hardware, Power, and Friction
4w ago00:11:01Tap to summarizeA lot of this week's news sounds separate until you line it up. Who owns the model, who controls the chips, who gets the data, and who absorbs the downside are all getting negotiated in public now. • Apple versus OpenAI • The new hardware map • Who owns the model • OpenAI and Anthropic tighten the screws • Economics, policy, and incentives • Markets and environmental constraints
Transcribe →Agents, Devices, and Data Trust
4w ago00:10:08Tap to summarizeA screen-free ChatGPT gadget sounds futuristic until you ask what it does better than the phone already in your pocket. And the same week we're hearing bigger promises about AI helpers, we're also seeing fresh reminders that data retention and oversight are still the part that can break everything. • OpenAI consumer device • AI data retention breaches • US cyber policy and model oversight • Human-AI collaboration skills • Agent engineering and governance
Transcribe →AI Fear, Hype, and Guardrails
4w ago00:11:08Tap to summarizeA lot of this week’s AI conversation comes down to a basic problem: people keep arguing from very different pictures of what AI is doing right now. Some are selling reassurance, some are selling urgency, and some are quietly looking at the data and finding a messier reality.Anthropic adPause petitions and what didn’t landA more grounded policy turnWhat the job market data actually saysKPMG and the real productivity bottleneckAI2040, AI2027, and the cost of acting on speculationHassabis and the fight over practical governanceThe Vatican enters the chatThis podcast was created with Podkey. Make your own at https://podkey.fm
Transcribe →AI Control and Cost Shifts
Jul 1300:09:59Tap to summarizeThis set of stories looks disconnected at first, but they all circle the same question: who gets to control AI when the stakes stop being theoretical. Governments are testing limits, labs are subsidizing adoption, and the hardware layer keeps looking more strategic by the week.Open-source AI policyUAE chip export waiverApple and OpenAI hardware fightToken subsidy warsEfficiency beats brute forceAI as reasoning partnerPrivacy backlash and memory demandThis podcast was created with Podkey. Make your own at https://podkey.fm
Transcribe →AI's Real Bottlenecks
Jul 1200:08:20Tap to summarizeA lot of companies say they're doing AI. Far fewer people are actually using it well, and some of the supposed productivity gains look a lot messier up close.The adoption gapWhat AI changes in the mindWhy work can feel worse, not lighterWhy internal champions matterWhere the big gains actually come fromWhat holds upThis podcast was created with Podkey. Make your own at https://podkey.fm
Transcribe →Curved Spacetime and Curved Incentives
Jul 1000:11:04Tap to summarizeToday’s mix is weirdly coherent. On one side, gravity stops being a force and becomes geometry. On the other, AI stops being a chatbot and starts becoming infrastructure, management theory, and maybe a governance problem.Why Einstein changed gravityGravity as geometry, not forceSchwarzschild, horizons, and what observers seeTime dilation, redshift, and why GPS caresBlack holes as energy machinesWhat counted as proofAI agents leaving the chat windowCheap models, shaky benchmarks, expensive chipsGovernance, consciousness, and who gets protectedAge gates and ambient surveillanceThis podcast was created with Podkey. Make your own at https://podkey.fm
Transcribe →