The Frontier

Jason and Lon Harris said Plaud records meetings on a phone or pin and turns audio into formats such as meeting notes, lesson plans, and interview Q&A. Jason said the show now uses Plaud to produce more comprehensive show notes than its former three-hour manual process.

Jason called Nathan Fielder’s Theranos documentary the greatest documentary he had seen, praising its layered treatment of Elizabeth Holmes, her husband Billy Evans, Silicon Valley’s fake-it-till-you-make-it culture, and responsibility for fraud. Holmes had been convicted and was filmed 30 days before incarceration; Jason said the film runs three hours.

Anthropic proposed putting 5% of its stock into a fund for public-benefit projects, subject to shareholder approval; its board could add 2% for each 50% increase in company value above an undisclosed baseline. Jason argued direct share distributions would give ordinary Americans a stake in AI’s gains, while Lon Harris questioned letting Anthropic leaders decide how the fund is spent.

Clone Robotics CEO Rad Krishnan said the company uses hydraulic artificial muscles instead of the electric motors and tendon-pulley systems common in humanoid robots. Pressurized fluid contracts a rubber tube inside a braided sleeve, producing linear motion; Krishnan said this approach could cost a fraction of motor-actuated designs while preserving human-sized strength and movement.

Krishnan said he began pursuing androids as a lifelong goal, built a one-megawatt plasma thruster at 13, and met his Polish co-founder on Facebook in 2020. He moved to Poland in 2021 to start Clone Robotics, after deciding that hydraulic muscle technology and newer deep-learning methods made the project viable.

Krishnan said Clone plans to deploy human-level torsos in enterprise settings before complete androids, with assembly-line and manufacturing work as early uses. He expects complete home robots in two to three years, described 2027 as a possible first deployment year, and cited $25,000 as a potential affordable price while also envisioning luxury models.

Krishnan said Clone is betting that open models and human-generated data will supply much of the robot’s intelligence, while the company focuses first on proprietary hardware. He expects models to turn over quickly and plans to adapt open-weight checkpoints with robot data, simulation, motion capture, and reinforcement learning.

Odyssey showed a simulated fruit-fly brain flying through worlds generated by its Odyssey 3 model, prompting discussion of whether a sufficiently realistic simulation could be indistinguishable from reality to its inhabitants. The hosts connected the demo to simulation theory and games such as Grand Theft Auto and Free Guy.

Jason called the chance of losing control of AI minuscule, estimating it at one in a billion, and argued that software would remain controllable. Lon Harris said the show should hear multiple sides, while Jason replied that AI-safety advocates already receive extensive media coverage; the hosts also debated religion and the ethics of creating simulated minds.

Jason said ElevenLabs’ Archer voice was based on podcaster Chris Williamson’s distinctive accent, despite ElevenLabs saying a different voice actor supplied it. The company removed Archer without admitting wrongdoing; Jason argued the resemblance could confuse listeners and urged ElevenLabs to compensate Williamson or license his voice.

Jason said founders seeking a premium domain should use a broker to conceal their identity and avoid bidding against their own public valuation. He cited GoDaddy’s $99.99 broker service and estimated clone.com could cost hundreds of thousands to millions; the hosts also joked that domain ownership disputes after an owner’s death could inspire a romantic comedy.

The hosts promoted startup services including Northwest Registered Agent’s free identity tools, Vanta’s compliance software, Rippling AI’s cross-system business dashboards, and GoDaddy’s domain and security services. Vanta said it serves more than 16,000 companies; the sponsors offered a 10% Plaud discount and $1,000 off Vanta.

Jason and Lon Harris said their Substack publishes tactical startup advice and uses Claude to develop infographic concepts, which Jason argued would be too slow and costly to produce with a human illustrator for every post. The team is hiring business writers across experience levels and plans to compare advice from people with opposing views.

Dylan Field says Figma, founded in August 2012, helps teams design and build software. Design is its core, but people also use Figma for marketing, advertising, presentations, whiteboarding, and other visual work.

Field argues that as AI makes coding easier, design, brand, marketing, and a clear point of view become stronger sources of product differentiation. He calls design the new code, while stressing that Figma still needs engineers.

Adobe agreed to buy Figma for $20 billion, but regulators in the United States, European Union, and United Kingdom blocked the deal after 16 months of review. Field says the review created costly uncertainty and that regulators may define markets too narrowly by looking backward.

Field says Roomba's blocked sale to Amazon was followed by bankruptcy and the market shifting to China, making the antitrust outcome counterproductive from a U.S. perspective. He uses the case to argue that regulatory ideals can have damaging practical effects.

When the Adobe deal ended, Figma had roughly 1,100 to 1,200 employees. Field says the all-hands brought relief because the 16-month ambiguity was over; he offered several months of pay to employees who wanted to leave, with reapplication allowed after six months.

Field says Anthropic's Claude Design did not significantly affect Figma, and the host cited Figma's 48% year-over-year growth after the announcement. Field argues that model progress is easy to confuse with product progress; professional design tools also require carefully built workflows.

Field rejects both blanket claims that SaaS is safe and that SaaS is dead. He points to Cursor's growth as evidence that a company can compete with frontier labs by obsessing over users, anticipating their needs, and building the right product before it has its own model.

Field says AI lowers the floor for producing digital work but does not raise the creative ceiling; people raise it by investing time and making bold choices. He says app counts keep growing while people use about the same number of apps, increasing the need to stand out through design.

Field says consensus-avoiding ideas can break through an attention market that quickly absorbs familiar points of view. At Figma, employees are asked to label AI-assisted writing so colleagues can judge whether the work has been reviewed and trust it accordingly.

Field recommends using AI to get past a blank page, then shaping the result against a human vision rather than accepting the model's default style. Figma aims to let users move between AI generation and hands-on editing; its Weave product connects model outputs and inputs in a node-based system.

Field describes design as creativity combined with problem-solving: designers must balance user needs, business goals, brand, culture, and the details of a product flow. He says human novelty-seeking and learning from consequences differ from current AI systems, while creative quality remains difficult to verify.

Field says an OpenAI release covering more than 700 papers on famous mathematical problems felt both exciting and sad: it showed major progress while making him feel that some mystery had been lost. He argues that mourning that loss can coexist with optimism about what the discoveries enable.

OpenCode said an ongoing, large-scale distillation attack exposed users’ API keys and affected 1.4% of console users; attackers used linked OpenAI, Anthropic, and Google accounts to generate costly API usage. OpenCode urged users to rotate keys, while OpenAI reset compromised keys and refunded affected customers.

OpenCode founder Dax Rad said frontier labs’ restrictions on cyber tasks initially hindered incident response, so his team relied on open-source models before gaining access to OpenAI’s Daybreak Blue program. The host framed the incident as a test of whether cyber defense should rely on lab-approved access or widely available models.

The host described the OpenCode breach as a possible first large-scale malicious attack by an agent swarm and compared its sustained activity to agents that overwhelmed Hugging Face. Stolen API access could be resold or used to generate synthetic training data, turning compromised services into directly monetizable assets.

Semiconductor stocks and the Nasdaq fell as investors questioned AI valuations, but the host said tighter financing for risky data centers did not yet show the demand collapse or canceled spending associated with the dot-com bust. Jensen Huang remained bullish, while Cathie Wood argued that fear of a bubble was itself evidence against one.

Microsoft canceled a planned Essex data-center contract after local utilities said grid connection might not arrive until the 2030s, despite the site originally targeting launch the following year. The host said the cancellation points to power and capital constraints, while Nscale supporters argued it does not signal weaker demand for capacity.