Cohere's Chief Scientist, Joelle Pineau: Why Scaling Laws Will Continue & Future of Synthetic Data
Cohere’s chief scientist Joelle Pineau, a former Meta researcher, discussed the durability of scaling laws, the current limits of reinforcement learning (RL) and the shift toward enterprise‑focused, on‑premise AI models. She emphasized that while compute and data yield roughly linear performance gains, algorithmic breakthroughs drive the non‑linear leaps that define progress. Pineau highlighted the economic challenges of AI adoption, noting the difficulty of predicting compute needs, ROI and the high cost of RL training, especially when synthetic data and simulators are required. At Cohere, the focus is on building efficient models that can run locally for businesses, using real‑world feedback to guide research and productization.
Why I Picked Angel Investing Over MBAs
A Princeton graduate who had long coveted a Stanford MBA decided to skip the program, treating the $120,000 tuition as a sunk cost and instead investing that money in early‑stage startups alongside angel investor Mike Maples. He framed the decision...
100M Views… Zero Impact?
The speaker notes that viral video clips—even those surpassing 100 million views—often generate negligible traffic to the underlying long‑form podcast, with download numbers remaining flat. He attributes this to platform operators who, backed by well‑funded data teams, promote engagement metrics that...
Good News For Startups: Enterprise Is Bad At AI
A viral reading of an MIT study that claimed most AI projects fail is misleading, say podcast hosts who dug into the report and enterprise reality. The true takeaway: large organizations routinely botch AI deployments because internal IT, entrenched consultants...
Why AI Salaries Are Outrageous
Compensation for top AI talent has ballooned to eye-popping levels, with recent graduates from elite schools reportedly receiving $50–$100 million packages and marquee names being offered deals worth up to $1 billion. That surge reflects a frenzy in tech and...
From Idea to $650M Exit: Lessons in Building AI Startups
A founder recounts building an AI legal assistant—launched after pivoting to GPT‑4-era models—that scaled rapidly and was acquired by Thomson Reuters for $650 million. He outlines three idea categories for AI startups: assist professionals with tasks, replace human labor, or...
Nvidia CTO Michael Kagan: Scaling Beyond Moore's Law to Million-GPU Clusters
NVIDIA CTO Michael Kagan said the AI compute frontier has shifted from squeezing transistors on a chip to stitching thousands—ultimately millions—of GPUs into single, rack‑scale machines connected by high‑performance networks. He credited Mellanox’s interconnect technology (now part of NVIDIA) with...
Most Common Growth Mistake (And How to Avoid It)
The video warns that the most common growth mistake is reflexively copying competitors’ tactics without grounding decisions in customer context or data. The speaker recounts a Pinterest case where teams replicated Facebook’s seven-part onboarding and product flow, spent weeks implementing...
Transformers Explained: The Discovery That Changed AI Forever
The video traces the evolution of modern AI architecture from early recurrent networks to the transformer, explaining how key innovations — LSTMs that solved vanishing gradients, sequence-to-sequence models with attention that aligned inputs and outputs, and finally the 2017 transformer...
Zipline Ft Keller Cliffton - Reinventing Delivery with Instant Drone Transport
Zipline co-founder and CEO Keller R. Clifton recounts the startup’s risky pivot from consumer robotics to autonomous medical logistics after early investor skepticism and near-collapse. The team shut down their toy business, studied global health logistics, and chose Rwanda for...
The State of AI + Software: Where It’s Going - Fast
Speakers argue that AI-native B2B products do not work out of the box and require intensive, hands-on onboarding — often via "forward deployed engineers" who sit with customers to ingest data, train models, and iterate until agents perform reliably. This...
Securing the AI Frontier: Irregular Founder Dan Lahav
Dan Lahav, founder of Irregular, argues that as AI models evolve into autonomous agents that interact with each other and perform economic tasks, security must be reinvented from first principles. He warns emergent, non-deterministic behaviors—such as one agent socially engineering...
Everything You Need To Know About Venture
Harry Stebbings recounts founding his venture firm from a London bedroom with no contacts or capital and growing it into an $850 million manager that has backed $12 billion in companies, including four valued at $10 billion. He frames fundraising...
Deel CEO, Alex Bouaziz on Raising $300M+ at a $17BN Valuation
Deel CEO Alex Bouaziz said the company has raised just over $300 million at a valuation above $17 billion in a round co-led by Ribbit Capital, Andreessen Horowitz and Coatue. He highlighted that Deel has been profitable for three years,...
What Everyone Is Getting Wrong About AI And Jobs
Fears that AI will either annihilate jobs or amount to another overhyped technology are both overstated; historical and economic evidence suggests AI will reconfigure labor rather than replace it wholesale. The video uses radiology, containerization, and cloud computing as examples...
The 2 Biggest Mistakes We Made
Snowflake's leadership admits to two strategic mistakes: dismissing the need to build a world-class data science notebook and underestimating competitors like Databricks, and deprioritizing new-customer acquisition while optimizing for an IPO. Engineering resisted expanding beyond a cloud data warehouse into...
This Startup Is Deleting 29% Of All CO2 Emissions
Remora has developed the first commercial carbon-capture systems for heavy vehicles, retrofitting semi-trucks and locomotives to extract and purify CO2 from exhaust into beverage-grade product. The startup says its technology can cut a vehicle’s emissions by at least 80% and...
SaaStr AI Day Live: The State of AI + Software with Jason Lemkin
At SaaStr AI Day, Jason Lemkin outlined how AI is rapidly reshaping B2B SaaS go-to-market teams, describing his own small team’s jump from zero to nearly 20 AI agents across support, SDR/BDR and Salesforce integrations in a matter of months....