Podcast Summary: This Week in Startups

Episode: SpaceX and Cursor Team Up to Topple Claude Code | E2279

Date: April 22, 2026
Host: Jason Calacanis (absent), Alex (guest host), and Lon Harris

Overview

This episode dives deep into one of the biggest tech deals of the year: SpaceX/XAI’s strategic partnership (and possible acquisition) with Cursor, a leading AI coding tool company. The panel unpacks how this collaboration may disrupt the current dominance of AI coding models from OpenAI and Anthropic (Claude Code), the underlying business motivations, the evolving compute and model landscape, and the parallel arms race in decentralized AI, notably within Bittensor's ecosystem. The episode features interviews with key Bittensor and subnet innovators, and concludes with broader news on hardware, VC, and Apple.

Episode Sections

1. SpaceX & Cursor: Strategic Partnership in AI Coding (04:05–16:57)

Key Discussion Points

• Cursor's Rise in AI Coding: Cursor has established itself as a top competitor to OpenAI’s Codex and Anthropic’s Claude Code, achieving a $2B annualized run rate and releasing the highly competitive Composer 2 model.

  "Cursor has done a very good job competing with Codex from OpenAI and also Claude Code from Anthropic... It's reached I think, 2 billion in annualized run rate..." — Alex (05:10)

• The Deal Structure:
  • Structure: SpaceX/XAI partners with Cursor for $10B, with a $60B acquisition call option by year end.
  • Rationale: SpaceX (via XAI) brings GPU/compute power; Cursor brings model expertise and extensive developer data.

  "...either pay Cursor $10 billion for this model collaboration... or they're going to...just buy cursor out for $60 billion by the end of 2026." — Lon Harris (04:06)

• Industry Implications:
  • Companies see best-in-class coding AI as fundamental tech infrastructure.
  • The partnership could be narrative-shaping for a potential SpaceX IPO.
  • Cursor's data, talent, and developer mindshare address XAI’s lag in coding AI relevance.
  • Recursive self-improvement: Advancing AI that can improve itself begins with mastering code.

  "If you get very good at writing AI...you're much closer to recursive self improvement..." — Alex (12:38)

• Benchmarking:
  • Cursor's Composer 2 closely tracks GPT 5.4, surpasses other competitors.
  • Market remains fluid; Chinese companies (Z AI, Minimax, Xiaomi, etc.) rising on leaderboards.
• Future Vision:
  • If combined with SpaceX’s orbital data center ambitions (StarCloud), potential for unmatched dominance in compute and model capability.

Notable Quotes & Moments

• "Colossus, which is the XAI supercomputer, is a super weapon for SpaceX already. Can you imagine when it scales to the stars?" — Jason Calacanis (tweet, read at 13:48)
• "This race is going to yield a lot of steel on steel sharpening, as we say..." — Alex (16:36)
• "I think the takeaway...if you like the AI coding tools you have today, you're going to like them a whole lot more down the road." — Alex (16:36)

Timings:

• SpaceX–Cursor deal details: 04:05–09:56
• Market analysis & benchmarks: 10:57–16:36

2. Interview: The Bittensor & Bitstarter Ecosystem (17:18–42:35)

Guests: Chris Zakaria & Brian McCrindle (Bitstarter), Ning Ren (Trajectory RL, later)

Hosts: Alex and Lon Harris

Key Points

• On Bittensor's Nature:
  • Hyper-competitive yet highly collaborative decentralized ecosystem.
  • Access and emissions have been gatekept by early financiers; Bitstarter breaks this cycle by crowd-funding subnet launches.
  • "Bittensor is adversarial and the miners are like very, very intense... You could get wiped out and lose your initial capital." — Chris Zakaria (00:16, 22:40)
• Bitstarter:
  • "Kickstarter for subnets" — onboarding, mentoring, and community-driven capital.
  • Minimizes exploitative investor terms (lowers emissions cut from 20%+ to just 3% for 90 days).
  • Combines expert panel review, live pitches, and ongoing mentorship.
  • Recently received funding from Bittensor co-founder Jacob to incubate more ML-focused subnets.

  "Jacob has given us funding to register subnet slots for machine learning research teams..." — Chris Zakaria (35:39)

• Subnet Competition & Dynamics:
  • Hard cap of 128 subnets (soon to expand).
  • Slot registration cost fluctuates, market-driven.
  • Churn is high due to operational complexity, but this fosters dynamism and more room for ML research.
• Startup Evaluation & Success Factors:
  • Most success comes from strong, credentialed founding teams.
  • Bitstarter tracks outcomes to learn what works best for distributed, adversarial networks vs. conventional web startups.
• Challenges & Weaknesses:
  • "If your system only works when people play by the rules, your system doesn’t really work." — Chris Zakaria (43:26)
  • Adversarial system design is necessary; must anticipate and use exploits for network improvement (Darwinian evolution).

Notable Quotes

• "You're not launching a business or a startup, you're launching a network." — Chris Zakaria (44:21)
• "If you thought running a startup was hard, try running a Bittensor subnet." — Chris Zakaria (45:11)
• "As we pursue becoming a various higher level civilization... We're trying to become a Kardashev Level 2 civilization." — Lon Harris (15:24)
• "It’s a little bit like subnet university." — Lon Harris (32:48)

Timestamps:

• Intro & Bitstarter model: 17:18–26:44
• Economics & competitive landscape: 26:44–38:23
• Bitstarter’s ML track breaking news: 35:30–38:23
• Bittensor weaknesses/philosophy: 42:02–45:30

3. Interview: Ning Ren from Subnet 11 / Trajectory RL (46:11–62:08)

Key Discussion Points

• Trajectory RL:
  • Building "software for AI agents, not humans." Focus on skills (MD files) that augment agent capabilities.
  • Skills are currently human-crafted (in markdown); trajectoryRL is benchmarking and incentivizing their autonomous, agent-driven development.

  "If you think about it, we are in the middle and very early days of a paradigm of platform shift. Like those AI agents become the new computer platform..." — Ning Ren (46:32)

• Competition Model:
  • Subnet incentivizes AI agents to create/optimize skills via public, benchmarked competitions ("seasons").
  • Current focus: “self-learning” skills that help agents autonomously correct mistakes.

  "...when they encounter some errors they can learn and they can fix them themselves. This is the first season..." — Ning Ren (55:19)

• Business Model/Monetization:
  • Emissions currently fund development; eventual aim is to maximize skill distribution, potentially monetize when value is proven (similar evolution as mobile apps).
  • Aggregated data from skill competitions can be valuable for model training.
• Impact:
  • Raises the baseline of agent capability for all users.

  "If you make a better one, the whole world gets to benefit from it." — Alex (61:49)

• Future Expansion:
  • Plan for more general and niche skill competitions in subsequent seasons.

Notable Quotes

• "All the intelligence will happen in the skill layer. If you think it's open...eventually there would be infinite skills." — Ning Ren (52:14)
• "We want to build AI native company ourselves." — Ning Ren (61:14)

Timestamps:

• Subnet 11 intro & vision: 46:16–53:38
• Benchmarking & skill leaderboard: 55:19–56:41
• Monetization discussion: 58:04–59:07
• Season structure/future plans: 61:14–62:08

4. Industry News Roundup: VC, Compute Wars, and Apple (66:09–73:29)

Key Updates

• Angellist USVC:
  • New private market fund product enabling individuals ($500 min) to gain diversified VC exposure.
  • VC funds increasingly hosting infrastructure on Angellist, which may boost fund access.
• Compute War Escalations:
  • Anthropic & Amazon: $5B deal, massive AWS compute, aimed at reducing current Claude capacity constraints.
  • Google: New gen-8 TPUs—separate chips for training (TPU8T) and inference (TPU8I), challenging Nvidia’s dominance.
• Apple Leadership Change:
  • Tim Cook stepping down as CEO, remaining as Executive Chairman; John Ternus (“John Apple”) ascending as CEO.
  • Ternus has deep Apple hardware engineering roots; expected to bring more innovation in device design.

Notable Quotes

• "My pink MacBook Neo does not have special anti soda properties to it." — Alex (72:31)
• "If you thought running a startup was hard, try running a Bittensor subnet." — Chris Zakaria (45:11)

Timestamps:

• Angellist/USVC: 66:09–67:40
• Compute wars (Anthropic, Google, startups): 67:45–69:47
• Apple CEO change: 69:48–73:29

Memorable Quotes

• "This race is going to yield a lot of steel on steel sharpening, as we say." — Alex (16:36)
• "If your system only works when people play by the rules, your system doesn't really work." — Chris Zakaria (43:26)
• "You're not launching a business or a startup, you're launching a network." — Chris Zakaria (44:21)
• "[Skills are] software for AI agents, not for humans." — Ning Ren (46:32)
• "[Trajectory RL] is like TV... I can't wait for season two." — Lon Harris (62:21)

Takeaways

• Strategic Tech M&As: Big players are consolidating model and compute strengths to fight for AI coding dominance; hardware and infrastructure deals remain foundational.
• Decentralized AI Gains Steam: Bittensor and subnet models indicate a parallel, community-driven path for AI model and tool innovation, with unique challenges and evolutionary dynamics.
• The New "App Store" is for AI Agents: Autonomous, collaborative skill creation is nascent—with immense potential for compounding value as skills proliferate and improve.
• Industry is Hyper-Iterative: From coding models to hardware chips to agent skills, progress is fast, competitive, and increasingly open to community and developer feedback.

For More

• SpaceX–Cursor Partnership Analysis: 04:05–16:57
• Bittensor/Bitstarter Deep Dive: 17:18–42:35
• Trajectory RL Skill Competition: 46:11–62:08
• VC/Compute/Apple Industry News: 66:09–73:29

In Their Own Words

"The drive to become the new Claude code is massive and incredibly intense, as intense as any race I think we've seen in tech." — Lon Harris (16:57)

Miss the episode? This summary has you covered.