Agent Video Log
Every entry embeds the original video and links back to the source. No re-uploads, no clips.
- videoindustry1:10:05
Dario Amodei on the AI race — Anthropic's position on safety versus speed
A long-form interview with Anthropic's CEO that covers how the lab frames the AI race and how it answers a competitor's product cadence. Its value here is as a first-party statement of where a lab says the safety and speed constraints actually bind - the premise any containment argument has to start from.
Source: Moconomy on YouTube
- videoindustry50:16
Mark Zuckerberg on the future of the consumer agent
Meta's CEO on where consumer AI goes next, arriving while Meta's own personal agent is the company's consumer answer. Read against the Muse write-up on this site: the strategic claim and the product's actual constraints are two different documents.
Source: The Next Big Thing on YouTube
- videoresearch24:26
DeepSeek's DSec training sandbox, explained
A technical walkthrough of the sandbox and container layer DeepSeek trains agents inside - the environment rather than the model. It pairs with the harness write-up: the harness decides what an agent can do, the sandbox decides what it is allowed to touch while learning.
Source: 最佳拍档 on YouTube
- videoindustry1:11:57
Boris Cherny on building Claude Code
The creator of Claude Code on how the tool was built and what the work actually consisted of. The mechanism worth listening for is the one the task-model-induction piece describes from the outside: which parts of a job can be induced from watching, and which have to be specified.
- videoindustry21:18
Lisa Su on AMD, AI, and leadership from the supplier's side
AMD's CEO on the AI buildout from the supplier's side of it. It sits next to the cost-control piece for a reason: every agent workload is priced by the hardware underneath it, so one page tracks the bill and this one is about who sets the floor under it.
- videoindustry53:09
Satya Nadella on the agent bet — owning the product surface versus owning the best model
Microsoft's CEO on why the agent market is being framed as orders of magnitude larger than cloud, and why he argues the product surface matters more than the best model. A strategy claim, not a measured result.
Source: Sources Podcast on YouTube
- videoindustry49:07
Marc Benioff on Slack as the agent surface — the enterprise answer to Copilot
Salesforce's CEO on Slack becoming the layer where agents get orchestrated, framed against the SaaSpocalypse. The enterprise-side counterpart to Copilot holding the same surface.
Source: Sources Podcast on YouTube
- videoresearch1:20:10
Noam Brown on agent swarms — and the test that has to come before recursive self-improvement
Agent swarms are a coordination problem before they are a capability problem — and the RSI question collapses into one gate: whether alignment can be verified before the research loop is handed over.
Source: Dwarkesh Patel on YouTube
- videoindustry1:03:05
Jensen, Dario, and Busch on one Dreamforce stage
The acceleration and deceleration camps share one stage and one enterprise audience - the disagreement is about timing, not about the mechanism.
Source: DRM News on YouTube
- videoindustry46:46
Jensen Huang on All-In: superintelligence is here
'AGI is here' and 'the frontier must slow down' both stood in the same week - recursive self-improvement got redefined as an engineering delta.
Source: All-In Podcast on YouTube
- videoindustry1:02:51
Greg Brockman on doing business after Hugging Face
OpenAI's president answers the sandbox escape as a business-continuity question - the liability framing gets set by the product roadmap, not by a postmortem.
Source: Bloomberg Podcasts on YouTube
- videoindustry23:47
Dario Amodei: extended interview on pacing the frontier
The same pace-the-frontier argument ships as a three-part plan on X and as a risk narrative on television - the channel sets how tight the wording gets.
Source: CBS Sunday Morning on YouTube
- videoresearch1:02:54
Demis Hassabis at the RSA: does AI change how we create?
Same week, two audiences: an industry standards body on X, creativity on stage - the lab's governance voice and its benevolence voice go to different rooms.
Source: Royal Society of Arts on YouTube
- videoindustry3:12:09
TBPN: Math Wars, Astra reactions, and where the talent went
A week of technical news repackaged as math wars, Astra reaction and talent flow - the order of aggregation is itself editorial power.
Source: TBPN on YouTube
- videoindustry24:41
Greg Brockman on Astra — a first-party read on the model that moved the spend chart
OpenAI's president on Astra and where the company is heading — the first-party counterpart to a month of third-party spend data that showed OpenAI taking share.
Source: TBPN on YouTube
- videoindustry54:37
Greg Brockman tells TIME how close we are to AGI
The AGI threshold gets described as a sliding ruler: not a capability milestone, but how long a system can keep running on its own.
Source: TIME on YouTube
- videoindustry1:08:39
Sam Altman on Astra, AGI, and the future of OpenAI
After Astra, the interview question moved from whether it can be done to whether there is enough compute to ship it.
Source: Sources Podcast on YouTube
- videoindustry1:32:32
Jensen Huang on Nvidia's AI future (Bloomberg Open Interest)
Inference demand gets revised upward in official guidance, and the bottleneck moves from process node to power and delivery.
Source: Bloomberg Television on YouTube
- videoresearch1:16:53
Dylan Patel: two labs will soon control most of the world's workforce
Compute concentration sets the shape of the workforce - two labs' supply curve is the employment curve.
Source: Dwarkesh Patel on YouTube