Skip to content
All-In PodcastAll-In Podcast

DeepSeek Panic, US vs China, OpenAI $40B?, and Doge Delivers with Travis Kalanick and David Sacks

(0:00) The Besties intro Travis Kalanick! (2:11) Travis breaks down the future of food and the state of CloudKitchens (13:34) Sacks breaks in! (15:38) DeepSeek panic: What's real, training innovation, China, impact on markets and the AI industry (50:14) US vs China in AI, the Singapore backdoor (1:01:51) OpenAI reportedly in talks to raise ~$40B with Masa as the lead investor (1:10:37) DOGE's first 10 days (1:25:13) Future of Self Driving: Uber, Waymo, Tesla (1:38:04) Fed holds rates steady, how DOGE can impact rate cuts (1:44:17) Fatal DC plane crash Follow Travis: https://x.com/travisk Follow the besties: https://x.com/chamath https://x.com/Jason https://x.com/DavidSacks https://x.com/friedberg Follow on X: https://x.com/theallinpod Follow on Instagram: https://www.instagram.com/theallinpod Follow on TikTok: https://www.tiktok.com/@theallinpod Follow on LinkedIn: https://www.linkedin.com/company/allinpod Intro Music Credit: https://rb.gy/tppkzl https://x.com/yung_spielburg Intro Video Credit: https://x.com/TheZachEffect Referenced in the show: https://github.com/deepseek-ai/DeepSeek-R1/blob/main/DeepSeek_R1.pdf https://www.tomshardware.com/tech-industry/artificial-intelligence/chinese-company-trained-gpt-4-rival-with-just-2-000-gpus-01-ai-spent-usd3m-compared-to-openais-usd80m-to-usd100m https://www.cnbc.com/2025/01/27/nvidia-sheds-almost-600-billion-in-market-cap-biggest-drop-ever.html https://x.com/shrihacker/status/1884414667503853749 https://x.com/balajis/status/1884975064283812270 https://www.fool.com/earnings/call-transcripts/2025/01/29/meta-platforms-meta-q4-2024-earnings-call-transcri https://x.com/mrexits/status/1885017400308806121 https://www.wsj.com/livecoverage/stock-market-today-dow-sp500-nasdaq-live-01-28-2025/card/deepseek-s-ai-learned-from-chatgpt-trump-s-ai-czar-says-LoCYvz2Lm0riS0AuEoB5 https://www.wsj.com/tech/ai/why-distillation-has-become-the-scariest-wordfor-ai-companies-aa146ae3 https://techcrunch.com/2024/12/27/why-deepseeks-new-ai-model-thinks-its-chatgpt https://x.com/rauchg/status/1875627666113740892 https://www.ft.com/content/a0dfedd1-5255-4fa9-8ccc-1fe01de87ea6 https://x.com/satyanadella/status/1883753899255046301 https://en.m.wikipedia.org/wiki/Jevons_paradox https://x.com/pitdesi/status/1883192498274873513 https://x.com/rihardjarc/status/1884263865703358726 https://x.com/austen/status/1884444298130674000 https://www.cnbc.com/2025/01/30/openai-in-talks-to-raise-up-to-40-billion-at-340-billion-valuation.html https://x.com/america/status/1884372526144598056 https://x.com/DOGE/status/1884396041786524032 https://fred.stlouisfed.org/series/FYFSD https://www.whitehouse.gov/presidential-actions/2025/01/establishing-and-implementing-the-presidents-department-of-government-efficiency https://x.com/Jason/status/1884671945800573018 https://abcnews.go.com/538/trump-starts-term-weak-approval-rating/story?id=118146633 https://www.cnbc.com/2025/01/15/cpi-inflation-december-2024-.html https://x.com/chamath/status/1885068981905875241 #allin #tech #news

Jason CalacanishostDavid FriedberghostChamath PalihapitiyahostTravis Kalanickguest
Jan 31, 20251h 49mWatch on YouTube ↗

CHAPTERS

  1. 0:00 – 7:10

    CloudKitchens, Ray Dalio, and the ‘Future of Food’ Vision

    The episode opens with banter about Friedberg’s surprise Ray Dalio interview before pivoting to Travis Kalanick’s return to the public stage. Kalanick outlines CloudKitchens as an infrastructure company—combining real estate, software, and robotics—to make high-quality, low-cost, ultra-convenient meals that rival grocery economics.

  2. 7:10 – 31:00

    Robotic Bowls, Nutrition Data, and Fully Wired Food Supply Chains

    The conversation dives into CloudKitchens’ robotics stack and how it connects to personalization, health data, and agricultural supply chains. Kalanick and Friedberg compare today’s automation efforts to early 20th-century Automats and past bowl-focused concepts like Eatsa.

  3. 31:00 – 43:00

    Sacks from the White House: DeepSeek R1 and the Global Freakout

    David Sacks joins from DC, describing the White House complex and quickly turning to DeepSeek R1’s release and Wall Street’s panicked reaction. He frames DeepSeek as a Chinese, open-source reasoning model that unexpectedly matched OpenAI’s o1, triggering both geopolitical and open-source-versus-closed debates.

  4. 43:00 – 55:00

    Debunking the $6M Training Myth and Following the GPUs

    Sacks challenges the viral narrative that DeepSeek achieved parity with Western frontier models for just $6M. He and Chamath dissect hardware estimates, export controls, and potential backdoors for NVIDIA chips into China, arguing that the real story is a massive, long-prepped compute cluster plus clever engineering.

  5. 55:00 – 1:06:00

    DeepSeek’s Technical Breakthroughs: GRPO, PTX, and Constraint-Driven Innovation

    The panel explores why DeepSeek’s engineering choices matter beyond cost: a new RL algorithm and low-level GPU coding that break from Western orthodoxy. They argue that constraints in China—limited top-tier GPUs and CUDA lock-in—led to optimizations Western teams may have overlooked because they had too much capital and compute.

  6. 1:06:00 – 1:24:00

    Did DeepSeek Distill OpenAI? IP, Cloud Security, and Open Source Fallout

    The hosts tackle the simmering allegation that DeepSeek heavily distilled OpenAI’s models—transforming ChatGPT outputs into training data for its own systems. They review behavioral evidence, OpenAI’s public claims, and the awkward position of cloud providers hosting potentially stolen IP while also selling access to the original vendors.

  7. 1:24:00 – 1:51:00

    Where Is the Value in AI? Shims, Apps, Data, and Mixture-of-Experts

    The besties shift from model drama to business strategy: if frontier LLMs are racing toward commodity status, where should founders and investors focus? They debate shims that abstract multiple models, data moats, application-layer businesses, and the likely rise of many small expert models over single giant ones.

  8. 1:51:00 – 2:18:00

    China, Copying, and the Evolution into an Innovation Powerhouse

    Kalanick recounts Uber’s China war as a case study in how ferocious copying morphs into genuine innovation. They then tie this to current Chinese leads in delivery, locker systems, and AI, and discuss whether export controls will slow or merely redirect China’s advance.

  9. 2:18:00 – 2:54:00

    OpenAI’s $40B Raise, Masa’s Style, and the Perils of Overcapitalization

    Rumors surface that OpenAI is raising $40B at a $340B valuation, possibly led by SoftBank’s Masayoshi Son. Travis draws on his experience competing with SoftBank-funded rivals to warn about both the power and risks of taking such capital, while the group questions whether sheer hardware scale is still a moat.

  10. 2:54:00 – 3:41:00

    Autonomy, Power Constraints, and the Coming Commercial Real Estate Shock

    Kalanick and the besties extrapolate what widespread autonomy and AI mean for cities, power grids, and real estate. They argue that the real choke points may be electricity and physical infrastructure rather than models or chips, and that parking-heavy land use could be radically disrupted.

  11. 3:41:00 – 4:16:00

    DOGE, Debt, and the Politics of Cutting $1–3 Billion a Day

    The focus shifts squarely to the Department of Government Efficiency (DOGE) and the Trump administration’s early moves to slash federal spending. The hosts connect buyouts, RTO mandates, and lease terminations to broader questions of deficits, interest rates, and the constitutional limits of executive power.

  12. 4:16:00 – 4:50:00

    Interest Rates, Treasuries, and Why DOGE Must Succeed Quickly

    The hosts link DOGE’s cuts to bond markets and interest-rate dynamics, arguing that fiscal credibility directly affects long-term yields. They warn that a move to 5.5–6% 30-year yields would be equivalent to double-digit rates on the early-2000s debt stock—economically devastating without swift action.

  13. 4:50:00

    Aviation Tragedy, Outdated Systems, and the Case for Automation

    The episode closes somberly with reflections on a recent DC-area aviation accident. The hosts relay feedback from commercial pilots and autonomy entrepreneurs about the risk profile at Reagan National (DCA) and the antiquated nature of U.S. air traffic control, arguing that modern software and automation could prevent such tragedies.

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.

Add to Chrome