03-12-2026Price:

The Frontier

Your signal. Your price.

AI & TECH

Democratized AI development outpaces public trust

Thursday, March 12, 2026 · from 3 podcasts, 6 episodes
This Week in StartupsThis Week in StartupsThis Week in StartupsPresidio Bitcoin JamTFTC: A Bitcoin PodcastTFTC: A Bitcoin Podcast
  • AI self-improvement is no longer theoretical; public tools like Karpathy's Auto Research enable non-experts to drive progress.
  • Crypto-incentivized networks like Bit Tensor monetize open-source contributions, creating a global, performance-based development market.
  • A massive adoption gap is forming: grassroots experimentation booms globally while US public sentiment remains deeply skeptical.

The ability to improve artificial intelligence is escaping the lab.

Andrej Karpathy’s Auto Research tool is a simple public loop. An AI agent rewrites its own code in five-minute cycles. Shopify CEO Tobi Lütke, not an AI researcher, used it over a weekend. His 37 experiments yielded a 19% performance gain on a small model. On This Week in Startups, Jason Calacanis argued this moves the field from a few thousand elite PhDs to potentially hundreds of thousands of practitioners.

Crypto networks are formalizing this shift into an economic model. Bit Tensor uses token emissions to subsidize 128 specialized AI subnets, paying developers globally to compete. Mark Jeffrey explained its coding assistant, Ridges, scores competitively with Claude but costs $29 per month. The project was built on roughly $10 million in chain emissions, while centralized competitors raised billions. This system monetizes open-source contribution, turning AI development into a performance-based contest.

Simultaneously, AI tools are lowering the barrier to build, not just research. On TFTC: A Bitcoin Podcast, Matt Corallo noted that tools like Claude 3.5 enable users to construct applications without deep coding knowledge. This democratization extends Bitcoin’s reach. Developers are also sorting tools into specialized roles. On Presidio Bitcoin Jam, DK described a triad: Gemini for code review, Claude for brainstorming, and OpenAI’s Codex as the relentless executor.

The global enthusiasm for these tools is not uniform. According to a poll cited by Calacanis, only 26% of Americans are pro-AI, with 46% opposed. This contrasts with places like China, where government-backed meetups for tools like OpenClaw draw massive grassroots interest. The technology is being built faster than public trust can form.

The dam has cracked. The question is who will control the flow.

Jason Calacanis, This Week in Startups:

- That 3,000 will turn into 300,000 people who understand how LLMs work and who can make meaningful progress on them.

- This is the dam cracking from the developers owning the world to everybody building the future.

AgentsBig TechOpen Source

Entities Mentioned

Google AntigravityProduct
StripeCompany
VisaCompany

Source Intelligence

What each podcast actually said

This Week in Startups
This Week in Startups

How agents will change banking forever | E2260 • Mar 10

  • Andrej Karpathy's Auto Research open-source tool proves AI models can already iterate and improve their own code within simple five-minute training loops.
  • Calacanis and Wilhelm note that while this isn't the full recursive self-improvement loop towards superintelligence, it is a working proof of concept for core autonomous improvement mechanics.
  • The tool's key impact is massive democratization. Shopify CEO Tobi Lütke, without an ML background, used it to run 37 experiments and find a 19% performance improvement in a small model over a weekend.
  • Jason Calacanis argues this shifts the landscape from a small elite of AI PhDs to hundreds of thousands of new tinkerers, moving 'from the developers owning the world to everybody building the future.'
  • Public tinkering experiments like these serve as a leading indicator that private labs at companies like OpenAI, Anthropic, and xAI are likely iterating at significantly faster rates.
  • The show's bullish prediction is that this acceleration sets up 2026 for potentially 'insane' rates of overall AI advancement and capability improvement.

Also from this episode:

Models (1)
  • Calacanis highlights a cultural split, noting Chinese governments are incentivizing AI adoption while a recent NBC poll shows only 26% of Americans are pro-AI, with 46% opposed.
AgentsModelsStartups

How agents will change banking forever | E2260 • Mar 10

  • Andrej Karpathy's Auto-Research tool enables an AI model to iteratively test and improve its own code in five-minute cycles, demonstrating a basic mechanic of self-improvement.
  • Shopify CEO Tobi Lütke used Auto-Research to run 37 experiments over eight hours, boosting a model's performance score by 19%, despite having no machine learning research background.
  • Calacanis argues that elite AI labs are likely advancing similar self-improvement techniques at a pace twice as fast as the public tools indicate.

Also from this episode:

Models (1)
  • Jason Calacanis predicts AI tool democratization will expand the pool of people capable of improving models from roughly 3,000 highly-paid PhDs to hundreds of thousands of tinkerers.
Society (2)
  • A recent NBC poll found only 26% of Americans view AI positively, with 46% opposed, indicating lagging public enthusiasm compared to technical progress.
  • The hosts contrast US skepticism with Chinese AI enthusiasm, where OpenClaw meetups draw crowds and local governments offer adoption incentives, driven by aspirational culture and tangible career utility.
Enterprise (1)
  • The barrier for non-technical executives to directly tinker with AI training loops has collapsed, foreshadowing tension with developers who prefer keeping management away from the codebase.
AgentsModelsBig TechStartups

Wisdom of the $TAO: the future is decentralized AI • Mar 6

  • The system monetizes open-source contribution in a way traditional development cannot, according to Jeffrey.

Also from this episode:

Mining (2)
  • Bit Tensor uses a crypto incentive layer with token emissions akin to Bitcoin mining rewards to subsidize AI development, according to guest Mark Jeffrey.
  • Jeffrey describes the model as Bitcoin's incentive structure applied to stranded talent instead of stranded energy.
Startups (7)
  • The network operates 128 specialized AI subnets that compete to produce the best models.
  • Ridges costs 29 dollars per month while centralized competitors raised funding at valuations in the billions.
  • The Ridges project was built on roughly 10 million dollars in chain emissions, compared to traditional startups requiring billion-dollar valuations.
  • Developers anywhere can earn subnet tokens daily by outperforming centralized teams, turning the stranded talent problem into a market.
  • A developer in Turkey can earn subnet tokens daily by improving the model, effectively owning a slice of the product's success.
  • The market bypasses traditional startup machinery including HR, payroll, and fundraising.
  • The network pays for progress directly, turning AI development into a performance-based contest.
Coding (2)
  • Subnet 62 launched Ridges, a coding assistant that scores 73 to 88 percent on benchmark tests measuring vibe coder effectiveness, according to Jeffrey.
  • Ridges scores competitively with Claude and Cursor on performance tests.
AI & TechMarkets
TFTC: A Bitcoin Podcast
TFTC: A Bitcoin Podcast

#723: The Battle for the Agentic Economy with Matt Corallo • Mar 8

  • Matt Corallo says the emerging agentic economy presents a major opportunity for autonomous AI payments, where agents will handle routine purchases like reordering household supplies, representing a genuine slice of future consumer spend.
  • Matt Corallo argues the race to build the default payment rail for AI agents is wide open, with entities like Google, Stripe, Visa, and crypto projects all pushing competing protocols from a starting point of zero.

Also from this episode:

Coding (3)
  • Matt Corallo argues that recent AI models like Claude 3.5 have crossed a threshold in the last three months, enabling the creation of functional software, from front ends to mobile apps, without human coding.
  • According to Matt Corallo, this leap in AI model quality removes the technical skill barrier for the Bitcoin community, allowing anyone with an idea and the will to execute to build Bitcoin applications.
  • Matt Corallo concludes that winning the agentic payment protocol war requires the Bitcoin community to step up and build, using the newly available AI tools to turn weekend ideas into working products.
Payments (2)
  • Matt Corallo states that legacy payment networks like Visa are useless for agentic commerce, as their systems are fundamentally anti-bot by design to prevent fraud.
  • Matt Corallo notes that stablecoins also fail to serve the agentic payment need due to a lack of merchant integration and usability for automated transactions.
Adoption (1)
  • According to Matt Corallo, this represents a unique shot for Bitcoin to achieve mainstream merchant adoption, as it is not trying to displace a 10x better incumbent but is competing in a newly forming market.
BTC MarketsPaymentsAdoption

#723: The Battle for the Agentic Economy with Matt Corallo • Mar 7

Also from this episode:

Models (1)
  • Matt Corallo says recent AI model advancements like Claude 3.5/3.6 have dramatically lowered the barrier to software development.
Coding (4)
  • He explains these AI tools now enable users to build robust frontend, web, and mobile applications without deep coding knowledge.
  • This marks a unique opportunity for the Bitcoin community, which thrives on experimentation and diverse builders.
  • Corallo says AI tools have eliminated excuses for Bitcoiners to build applications.
  • He says the tools exist for building, and now willpower and a clear concept are the only requirements.
AI & Tech (2)
  • The other major shift is the rise of 'agentic payments' where AI agents autonomously purchase goods and services.
  • Corallo states this isn't a distant future and will soon comprise a non-trivial portion of consumer spending.
Markets (3)
  • Existing payment rails like traditional credit card sites are not equipped for agentic payments, as they employ anti-bot measures.
  • Traditional systems also struggle with chargeback structures designed for humans, not autonomous agents.
  • For agentic payments, Corallo argues everyone is starting from zero, creating a greenfield opportunity.
Stablecoins (1)
  • Stablecoins face a similar hurdle, lacking widespread merchant integration for agent-to-merchant transactions.
AI & TechAdoption
Presidio Bitcoin Jam
Presidio Bitcoin Jam

Codex vs Claude Vibe Coding, Study Shows AI Agents Prefer Bitcoin, Kazakhstan to Add BTC? • Mar 7

Also from this episode:

Coding (9)
  • Developer DK claims OpenAI's Codex CLI has overtaken Claude Code for execution-heavy tasks, describing Codex as the relentless "builder" and Claude as the "brainstormer".
  • DK advocates for a three-tier AI coding workflow using Google's Gemini for code review, Anthropic's Claude for architecture exploration, and OpenAI's Codex for persistent execution.
  • DK previously relied on Claude Code for months but found it gets stuck in rabbit holes when exploring ideas like an artist, whereas Codex focuses like "a dog on a bone" through refactoring tasks.
  • Developer Callie characterized Claude as working like an "American" and Codex like a "German" in their respective approaches to software development.
  • DK conducted a "vibe coding" session at 70 miles per hour through the Nevada desert using Tesla's Full Self-Driving to handle highway driving while simultaneously using OpenAI's Codex CLI for software architecture.
  • The desert coding setup involved speaking commands to the terminal, letting the AI process for ten-minute intervals, and checking the screen periodically over a five-hour period.
  • Grok has stagnated as a competitive coding assistant over the past six months despite its integration with Tesla vehicles, according to DK.
  • Tesla's Grok integration allows drivers to hold the steering wheel button to speak commands and later receive code on their laptop, functioning as a car convenience rather than a serious coding contender.
  • DK described Codex as "like your autistic friend who just keeps going" and stated it is "insanely better than the alternatives right now at this moment."
Safety (1)
  • Tesla's Full Self-Driving capability enables "vibe coding at 70mph," which raises safety concerns about using AI to write code while AI operates a vehicle at highway speeds.
AI & TechCoding