Software moats vanish as AI agents commoditize code

AI & Tech (3)

• Robert from Manifold explains Targon uses Bit Tensor's Subnet 4 to aggregate encrypted GPU compute via a confidential virtual machine, leveraging TDX, AMD SEV, and Nvidia confidential compute to secure data on permissionless hardware.
• Targon acts as a buyer of last resort for data center GPU capacity, currently sold out, with pricing visible on stats.targon.com and plans to transition from an auction to an orderbook system for market-based rates.
• Jason Calacanis launched a $5,000 bounty via annotated.com to build a service for clipping and commenting on text, video, or podcast snippets, creating threaded discussions and fact-checks to train LLMs.

Protocol (1)

• Jason Calacanis argues Bitcoin's relevance is fading as stablecoins dominate payments, developer activity shifts to platforms like Bit Tensor and Solana, and incremental buyer demand weakens without new utility beyond speculation.

Business (1)

• Alex Wilhelm reports cloud capex surges continue, with Google Cloud revenue up 63%, AWS growth hitting 28% - its best in 15 quarters - and Microsoft, Amazon, and Meta all increasing planned spending for 2025.

Politics (1)

• Jason Calacanis highlights the political risk for US startups using Chinese open-source AI models like Qwen or DeepSeek, citing congressional pressure on companies like InSphere and Cursor, though he views backdoor threats in open models as limited.

AI Infrastructure (2)

• Magnific integrates third-party state-of-the-art models, including from Google, alongside proprietary upscaling and skin-enhancer models, to provide users with the best available creative output.
• Joaquin argues AI video tools raise both the creative ceiling and floor, enabling projects that were previously too expensive to get greenlit while also empowering smaller teams and individual creators.

AI & Tech (3)

• Gal Gadot told Jason Calacanis that AI tools allow film productions to cut costs by two-thirds, letting actors focus on performance and enabling a potential Cambrian explosion of new content.
• Joaquin believes AI will match the creativity of some humans but cannot replicate human individuality, predicting increased demand for people who can inject their unique experiences and storytelling into projects.
• Joaquin notes that while AI can already localize static ads across languages and cultural details, generating hundreds of localized video variants remains error-prone and requires better steering and validation systems.

AI & Tech (1)

• Critic Dwayne OnX argues OpenAI’s GPT-5.4 fails at UI design and lacks aesthetic judgment, requiring explicit design file inputs to produce acceptable work.

Models (1)

• OpenAI's GPT-5.4 is now available as a limited preview on AWS, with GPT-5.5 coming soon, a direct result of the amended partnership ending Microsoft's exclusivity.

Big Tech (1)

• He gives Google a low momentum score of 3 out of 10, citing its struggle to break into the agentic and coding-led conversation of 2026, despite strong narrative positioning at the year's start.

AI & Tech (1)

• He cites analyst Miles Brundage to counter zero-sum thinking, arguing the rapidly expanding AI pie means multiple labs can succeed, with Semi Analysis's Dylan Patel noting token demand outpaces supply.

AI Infrastructure (1)

• Whittemore cautions that data from the pre-agentic era, like a WSJ report on OpenAI missing revenue targets, is now a lagging indicator and won't reflect the current structural shift in the industry.

Chips (1)

• Google Cloud unveiled its eighth generation of TPUs (8T for training, 8I for inference), offering three times faster training performance and 80% better performance per dollar.

Models (1)

• OpenAI released GPT 5.5, seven weeks after GPT 5.4, showcasing a 37-point increase in long-context reasoning and a 60% reduction in hallucinations compared to 5.4.

AI & Tech (6)

• Alex reports that Frontier Math Tier 4 benchmark, a proxy for professional-level math, shows approximately 1% monthly gains from frontier AIs, suggesting all such problems could be solved in four to five years.
• The trial between Elon Musk and OpenAI in Oakland Federal Court has begun with jury selection, unveiling details through discovery that are likely to be publicly aired.
• Dr. David Lutske's viral post demonstrates Grok creating a realistic AI Frenchwoman with a reflective ID, suggesting video ID verification may soon be unreliable.
• OpenAI released ChatGPT for clinicians, a free AI co-pilot that outperforms human doctors on health benchmarks, scoring 59 compared to 43.7 for human clinicians.
• Project Top Heart by NYU and Stanford uses AI to analyze 20 variables, aiming to increase the number of viable donor hearts for transplant by an additional 500 per year.
• The FDA-approved blood pressure medication Candesartan has been repurposed using AI to stop and inhibit MRSA infections, which affect 2.8 million people and kill 35,000 annually in the U.S.

Science (2)

• MRNA vaccines for pancreatic cancer show lasting results, with 87.5% of patients who generated a strong immune response still alive after six years, compared to a historical 13% survival rate.
• A single-shot CAR-T infusion achieved 100% cancer-free status in all 20 melanoma patients within two months of treatment, with no recurrence after a median follow-up of 15.3 months.

AI & Tech (3)

• Russ D'Sa's LiveKit provides infrastructure for agents to see, hear, and speak. OpenAI used LiveKit's commercial product to build ChatGPT Voice, which exposed LiveKit's technology to hundreds of millions of users.
• DeepSeek-V4 was cited as an example of a high-performing, low-cost open-source model, with DeepSeek-V4 Pro priced at $348 per billion output tokens compared to Claude Opus 4.6's reported cost of $25 million per output token.
• Matan Grinberg states that the United States has fallen behind China in open-source AI innovation, which he finds embarrassing, and that China has a huge talent advantage, with Jensen Huang noting a majority of the world's best AI researchers are Chinese.

Protocol (2)

• Paolo Ardoino says Tether holds over 130,000 Bitcoin and is a continuous buyer.
• The WDK library provides self-custodial wallet creation for people, machines, and AI agents, with built-in support for Bitcoin's Lightning Network for scale.

Stablecoins (3)

• Ardoino frames Tether's mission as creating stability against societal 'darkness' marked by war, inflation, and currency instability, inspired by Asimov's psychohistory.
• Tether operates in 160 countries with 573 million users across its USDT, Tether Gold, and other services, adding 34 million new wallets per quarter.
• Ardoino introduces the 'resilience stack', Tether's open-source infrastructure suite designed to outlast societal instability, with over 1,000 projects on GitHub.

AI & Tech (4)

• Ardoino cites 4 billion people globally lack access to basic financial services, creating a societal gap that AI will widen 100x for the excluded population.
• Holepunch is Tether's rebuilt, cryptographic peer-to-peer protocol for scalable real-time data, enabling server-less applications like a peer-to-peer Uber.
• Keet is Tether's messaging app built on Holepunch, with over 5 million users, designed to be unstoppable and will be fully open-sourced.
• Key Vault is Tether's open-source SDK for building private, local AI tools, applying a 'not your keys, not your coins' principle to artificial intelligence.

Models (6)

• Andrej Karpathy defines software 1.0 as explicit rules, software 2.0 as learned weights, and software 3.0 as programming via prompting and the LLM context window as a lever over an interpreter.
• Karpathy says his MenuGen app, which uses OCR and an image generator to illustrate menus, is rendered obsolete by software 3.0. The raw approach is to give a menu photo to Gemini with NanoBanan and get a directly annotated image.
• Karpathy posits that future computing could invert the current architecture. Neural networks would become the host process, with classical CPUs serving as co-processors for deterministic tasks.
• Karpathy's verifiability framework holds that LLMs excel in domains where outputs can be verified, like code and math, because frontier labs use reinforcement learning with verification rewards during training.
• Karpathy cites the 'car wash' problem as current jaggedness: state-of-the-art models can refactor a 100k-line codebase but incorrectly advise walking 50 meters to a car wash.
• Karpathy notes that GPT-4's chess capability improved significantly from GPT-3.5 not just from scaling, but because a large amount of chess data was added to its pre-training set.

AI & Tech (2)

• Karpathy describes current infrastructure as built for humans, not agents. His pet peeve is documentation that tells a human what to do instead of providing text to copy-paste directly to an agent.
• Karpathy endorses a tweet stating 'you can outsource your thinking but you can't outsource your understanding.' He sees LLM knowledge bases as tools to enhance, not replace, human understanding.

Coding (2)

• Naval built a personal app store that lets him oneshot custom apps like a workout tracker, which then appear on his phone. He notes Apple's device keying prevents wide distribution but allows apps for friends and family.
• Vibe coding expands software creation from 0.1% of the population to maybe 3%, Naval estimates. It requires a clear vision and basic computer understanding, but eliminates team compromises and activation energy.

AI & Tech (3)

• Coding is easier to train AI on than creative writing because it offers vast data and easy verification through compilation and tests. Domains with sparse data or subjective quality, like creative writing, remain human opportunities.
• State-of-the-art context windows are about one million tokens, but as codebases grow, models lose the plot. This forces the human operator to guide architecture and debugging, preventing hacks and preserving features.
• Naval uses different AI models for different strengths: Claude for visual artifacts and meeting his level, ChatGPT as the all-around OG, Gemini for search and YouTube access, and Grok for unneutered truth and technical problems.

Coding (2)

• Mario Zechner auto-closes all first-time pull requests to filter out AI-generated spam. His GitHub workflow posts a comment asking for a human-written issue; agents ignore the comment, but humans respond, earning future PR privileges.
• Armin Ronacher warns the industry's 'dark factory' approach of deploying armies of agents with vague specs will produce low-quality software. The output quality is bounded by the mediocre training data the models use to fill specification gaps.

AI & Tech (1)

• Both hosts argue the real value of AI agents is automating tedious work to free up human time for design and polish, not maximizing token output. They say the current hype pushes for unsustainable speed at the cost of quality and engineer well-being.

Business (5)

• A barrel of oil is a 42-gallon unit of measurement. The global market consumes roughly 105 million barrels daily.
• Rory Johnston states roughly 20 million barrels per day transited the Strait of Hormuz before its closure. Current shut-in volume due to the closure is estimated at 13 million barrels per day.
• The oil futures curve signals market tightness through backwardation. A record-high prompt spread of $15 for WTI created a massive incentive to sell barrels immediately.
• Johnston argues the market is underreacting to the supply shock. His model suggests Brent could approach $200 per barrel by late June if the Hormuz closure persists and draws down OECD stocks.
• The dominant market narrative shifted from 'peak oil supply' fears in the 2000s to 'peak oil demand' driven by shale technology and the energy transition.

Energy (5)

• Spare production capacity is held almost exclusively by state actors like Saudi Aramco. The U.S. has no meaningful spare capacity due to its private, competitive industry structure.
• Not all crude oil is equal; value depends on density and sulfur content. Light, sweet crudes like Brent and WTI are more valuable than heavy, sour grades like Western Canadian Select.
• The economic impact of high oil prices is a regressive tax, hitting poorer consumers hardest. Demand destruction often comes from recession-induced income loss, not direct price elasticity.
• Johnston believes the Hormuz crisis will end when market pressure forces a U.S. concession. He notes a paradox where Trump's verbal interventions lower prices, temporarily reducing that pressure.
• While the U.S. is energy secure, coastal consumers still face global prices. Johnston cites literature showing U.S. presidential approval ratings move inversely with pump prices.

Politics (1)

• The long-term consequence of the Hormuz crisis will be accelerated energy transition investment, shifting the debate from climate morality to energy security and affordability.

BTC Markets (2)

• David Hoffman notes that while many cycle investors on Bankless remain bearish on crypto's short-term bottom, Jordi Visser holds a bullish outlook, believing Bitcoin has already bottomed and that the current crypto winter will be the mildest ever.
• Jordi Visser describes Bitcoin's recent price action as an 'IPO' event, involving a significant distribution from early holders to new buyers, including ETFs, which have continued to accumulate during price dips.

AI & Tech (3)

• Jordi Visser predicts that artificial intelligence and inflation will drive investors toward a 'scarcity portfolio,' ultimately concluding with Bitcoin and other assets possessing similar properties, due to a massive economic transition.
• Jordi Visser argues that AI accelerates wealth distribution problems, which have grown since the personal computer era, by disrupting human intellect and physical labor, making Bitcoin an inevitable and chosen scarcity asset in this new paradigm.
• Jordi Visser states that AI acts as the new quantitative easing (QE), enabling companies to reduce labor while growing, contrasting with traditional QE which aimed to keep businesses alive by maintaining credit flow.

AI Infrastructure (2)

• Jordi Visser identifies a 'compute shortage' as a critical current issue, as AI adoption rates have outpaced the supply of data centers and necessary hardware, potentially slowing companies' ability to replace labor and impacting margins.
• Jordi Visser's portfolio is heavily weighted towards 'scarcity assets' supporting the AI infrastructure, including memory stocks like Micron and Pure Storage, chip-related companies like Marvell, and raw material producers like silver miners and Brazilian mineral companies.

Macro (2)

• Jordi Visser forecasts a period of inflation driven by underinvestment in physical infrastructure like power and chips needed for AI, alongside rising commodity prices for copper, silver, and energy, despite AI's long-term deflationary potential.
• Jordi Visser notes that year-over-year CPI is currently 3.3%, predicting it will reach 3.6% or higher after the next print in early May, potentially surpassing 4% due to filtering effects from rising diesel and plastic prices.

Protocol (2)

• Jordi Visser observes that the current Bitcoin cycle is distinct from previous ones because altcoins have not reached their 2021-2022 highs, suggesting a reshaping of the crypto market reminiscent of the post-dot-com bubble era.
• Jordi Visser asserts that Bitcoin's strongest historical performance, with annualized returns of 247%, occurred when year-over-year CPI was above three-month bills and the Fed was on hold or easing, a regime he believes the market is rapidly approaching.

Education (1)

• He advises students to master AI as a powerful toolset and apply it to their field of interest, comparing its transformative potential to electricity.

Politics (1)

• Horowitz says tech had little voice in Washington under Biden, citing policies that nearly ended the crypto industry and an executive order requiring pre-approval for global GPU sales.

AI & Tech (1)

• He argues the biggest AI danger is over-regulation and a U.S. failure to compete, which could lead to China achieving superintelligence first, creating a dangerous power imbalance.