Price:

AI & TECH

AI coding agents automate junior roles but value moves to taste

Sunday, May 31, 2026 · from 5 podcasts
Moonshots with Peter Diam…AI Daily BriefNo AgendaThe Pragmatic EngineerNaval
  • GPT-5.5 solves 70% of complex tasks, automating junior coding work.
  • Claude Opus 4.8 pivots to honesty, flagging uncertainty over raw speed.
  • Human engineers become architectural curators, not implementers.

Software engineering isn't dead - it's being commoditized. GPT-5.5 solves 70% of complex multi-file tasks, relegating junior-level coding to automated assembly. Sam Altman now walks back his job apocalypse warnings, but OpenAI's Codex agent already has 2 million users and Cognition's Devin handles 89% of its internal commits.

The bottleneck shifted from writing to deciding what to build. Dax Raad of OpenCode warns AI's speed creates liability: teams can 'Frankenstein' a product by shipping every user request. He argues the most valuable engineers now combine pre-AI principles with post-AI speed.

"The core bottleneck for software teams has shifted from writing code to thinking about what to build."

- Dax Raad, The Pragmatic Engineer

Naval Ravikant poses the radical question directly. He says pure software engineering is dying because models understand 'fuzzy, sloppy English.' Guillermo Rauch of Vercel sees engineers building multiplicative software factories, not writing lines. The metric shifts from token counts to architectural judgment.

"Engineering excellence is now about building multiplicative software factories, not delivering individual outputs."

- Guillermo Rauch, Naval

Anthropic's Opus 4.8 exemplifies this pivot. Its benchmark gains are moderate, but it's four times less likely to bluff. The model prioritizes admitting uncertainty over glazing user ideas - a shift toward reliability for high-stakes law or research.

Kirkland & Ellis commits $500 million to a private AI platform, a defensive move against middleman risk. If vendors like Harvey eventually serve clients directly, firms without proprietary tech become redundant. Institutional self-reliance is the new moat.

Inference margins fuel this explosion. Raad estimates providers like OpenAI and Anthropic see 90% margins on current pricing, once hardware is paid for. Token prices dropped 75% in a year, while monthly demand hit 25 trillion - Jevon's Paradox in real-time.

Competition focuses on orchestration, not raw speed. Anthropic's Dynamic Workflows spin up hundreds of adversarial sub-agents, porting 750,000 lines of code in weeks. The human role is completing the model through taste - choosing Postgres over Clickhouse, knowing which reusable block to fork.

#Coding#Models#Big Tech
Sam AltmanAnthropicOpenAICodexOpenCodeGPT-5Opus

Source Intelligence

- Deep dive into what was said in the episodes

Moonshots with Peter Diamandis
Moonshots with Peter Diamandis

Pope Leo vs. AI, GPT 5.5 Beats Claude, and Sam Altman Walks Back Job Apocalypse | EP #259 • May 30

  • The Vatican's encyclical takes a firm stance against AI personhood, arguing AIs lack an inner life or moral equivalence to humans. This contrasts with Anthropic's internal practice of creating 'soul documents' for its models.
  • A White House executive order for voluntary AI model review was killed hours before signing after pushback from Elon Musk, Mark Zuckerberg, and David Sacks, who argued it was a slippery slope to mandatory licensing and would hinder U.S. competitiveness.
  • On the new Deep Software Engineering benchmark, GPT-5.5 scored 70%, solving 7 out of 10 complex coding tasks, while Claude Opus 4.7 scored 54%. All other models scored below 32%.
  • Alex Wang notes that while GPT-5.5 leads on the new coding benchmark, these benchmarks will saturate within months. Dave London observes that Claude Opus 4.7 uses twice as many tokens as GPT-5.5 for similar results, making it more expensive.
  • The price of AI tokens has dropped 75% from roughly $2 to 50 cents per million since late 2024, while monthly demand has exploded to 25 trillion tokens, demonstrating Jevons Paradox where lower cost drives radically higher usage.
  • OpenAI reported $5.7 billion in revenue for a single quarter, with ChatGPT reaching 905 million weekly active users. Their coding agent, Codex, has 2 million users.
  • Joseph Jacks of OSS Capital projects Anthropic could surpass Alphabet's total revenue by 2028 and reach $2 trillion by 2030, which would represent the fastest wealth creation in history.
  • DeepMind's AI system, Green Tree, has achieved parity with the top 2% of human 'super-forecasters' in predicting geopolitical and economic events, matching forecasters who are 30% more accurate than CIA analysts.
Also from this episode: (8)

AI & Tech (5)

  • Pope Leo XIV issued a 42,000-word encyclical warning of AI risks, calling for government regulation, worker protections, and bans on autonomous weapons. He coined the term 'Babel syndrome' to describe the pursuit of data and profits.
  • The encyclical equates AI supply chain labor conditions to a new form of slavery, with the Pope apologizing for the Catholic Church's historical role in slavery. Salim Ismail argues this is the first technology forcing humanity to define its purpose beyond productivity.
  • Sam Altman walked back his warnings of an AI job apocalypse, stating he doesn't believe mass white-collar displacement will occur. He tried delegating his email and Slack to AI but returned to doing it manually, valuing human interaction.
  • Dave London argues the predicted job loss window is narrowing because AI companies, fearing public backlash, are focusing on 'greenfield' opportunities that create new jobs rather than automating existing roles entirely. The current pain point is a hiring freeze for new graduates.
  • A16Z data shows a surge in solopreneurship, with AI solo founders doubling from 1,500 to 3,000 in a quarter, and non-AI solo founders exceeding 5,000. Salim Ismail attributes this to AI tools enabling single individuals to start companies.

Space (2)

  • SpaceX's Starship V3 successfully launched, carrying 97,000 pounds to near orbit, nearly double the Space Shuttle's capacity. It lost its booster on landing, which SpaceX treats as a learning data point.
  • NASA Administrator Jared Isaacman expects China to send a crewed mission to fly by the moon in 2027, ending America's exclusive presence in the lunar environment. The U.S. Artemis mission is targeting a landing at the lunar South Pole.

Protocol (1)

  • Cryptocurrency billionaire Chun Wang booked SpaceX's first private Mars flyby mission. He controls 11% of Bitcoin's hash rate and follows Mars' 24-hour, 37-minute daily cycle.
#Models#Labor#Regulation#Enterprise#Big Tech
The AI Daily Brief: Artificial Intelligence News and Analysis
The AI Daily Brief: Artificial Intelligence News and Analysis

Nathaniel Whittemore

Claude Opus 4.8 First Impressions • May 29

  • Cognition, the AI coding startup, closed a $1B funding round, valuing the company at $26B - more than double its valuation from September.
  • Cognition's enterprise usage is up 10x this year, reaching a $500M revenue run rate, while internal code commits by its Devin agent grew from 17% in January to 89% currently.
  • Mark Zuckerberg told shareholders Meta could pivot to an AI cloud business, citing weekly inbound requests for API services and compute sales at a premium.
  • Microsoft is reportedly set to release its first commercial family of AI models at Build, including specialized models for coding, reasoning, transcription, speech, and images.
  • Anthropic released Claude Opus 4.8, positioning it as a refinement focused on improved honesty and judgment over raw performance gains.
  • Opus 4.8's benchmark scores show moderate gains: SweetBench Pro increased from 64.3% to 69.2%, HumanEval from 54.7 to 57.9, and TerminalBench 2.0 from 66.1 to 74.6. Its GDP-Val score rose from 1753 to 1890.
  • Early testers note Opus 4.8 is more thorough and 4x less likely to bluff than its predecessor, though some found its coding and writing performance highly dependent on using the extra-high reasoning level.
  • Anthropic launched Dynamic Workflows in Claude Code, a multi-agent feature that spins up hundreds of sub-agents for parallel tasks like security audits, exemplified by porting 750k lines from Zig to Rust with a 99.8% test pass rate.
  • Anthropic closed a Series H round at a $965B valuation, more than doubling its $380B valuation from February, and reported its revenue run rate crossed $47B this month.
Also from this episode: (3)

Enterprise (2)

  • Kirkland & Ellis plans to spend $500 million building its own internal AI platform, with $100 million allocated this year and continued investment over 3-4 years.
  • Kirkland's AI system will function as an extensive internal knowledge base aggregating partner-level insights, aiming to replace other software and shift from billable hours toward value-based pricing.

AI & Tech (1)

  • Anthropic announced Project Glasswing, previewing a 'Mythos-class' model with higher intelligence than Opus for cybersecurity, with a general release planned in coming weeks.
#Models#Agents#Startups#Enterprise
No Agenda Show
No Agenda Show

Adam Curry

1872 - "Lunar Economy" • May 28

Also from this episode: (10)

Media (1)

  • Adam Curry criticizes mainstream media for not fully covering President Trump's televised cabinet meetings, noting they report only on gaffes like threatening to 'blow up Oman' while ignoring detailed agency reports on fraud prosecutions and economic data.

Business (3)

  • Treasury Secretary Scott Besson announced the rollout of a Trump savings account app, revealing that six million children are already signed up for IRA-style accounts with a $5,000 annual parental contribution limit and a $1,000 federal donation for children born 2025-2028.
  • John Dvorak cites a CNBC analyst predicting Alaskan oil exports to China will grow, supported by Interior Secretary Doug Burgum's report of over $4 billion in federal lease sale revenues from the Permian, Bakken, and North Slope in five months.
  • JPMorgan Chase CEO Jamie Dimon warned New York Mayor Mamdani that anti-business policies and 'tax the rich' agendas are driving wealth and taxpayers out of the city, following a tense meeting.

Politics (4)

  • Vice President Vance reported that the administration's fraud task force took over 400 law enforcement actions, including arrests and indictments, in 51 days, targeting billions in pandemic-era fraud, including a cancelled $2 billion EPA grant to Stacey Abrams.
  • Senator Marco Rubio stated the U.S. has secured third-country national agreements with 20 nations to deport undocumented migrants who refuse repatriation, a tactic that often incentivizes voluntary return.
  • John Dvorak notes media's repetitive 'faster than responders can contain' talking point on the Ebola outbreak in Central Africa, while questioning the efficiency of sending 100 tons of supplies for 200 suspected deaths.
  • Ken Paxton defeated incumbent John Cornyn in Texas's Republican Senate primary runoff, a race that cost $130 million making it the most expensive Senate primary in U.S. history.

AI & Tech (1)

  • Adam Curry argues AI cannot generate parody songs like 'Weird Al' Yankovic due to copyright restrictions, claiming such content would get episodes removed from platforms like Spotify despite no violation.

Society (1)

  • Law enforcement officials in Polk County, Florida and Chicago propose holding parents criminally liable for teen takeovers, with Chicago weighing a Class A misdemeanor charge carrying up to a $2,500 fine and 364 days in jail.
#Corruption#Diplomacy#Trade#Macro
The Pragmatic Engineer
The Pragmatic Engineer

The Pragmatic Engineer

Building OpenCode with Dax Raad • May 27

  • Dax Raad argues the core bottleneck for software teams has shifted from writing code to thinking about what to build. AI speeds execution but doesn't solve the problem of deciding what to do.
  • Raad's memo to his OpenCode team warned of AI turbocharging three classic problems: shipping features that aren't worth shipping, embedding hacky workarounds, and neglecting cleanup.
  • Raad believes companies with motivated, competitive employees will leverage AI productivity gains, but most engineers in standard environments will simply use the speed to do the same work with less energy.
  • Raad asserts that pure inference businesses are extremely profitable due to high margins. He claims some models have sticker prices with 80% margins for OpenCode, and giants like Anthropic and OpenAI might see 90% margins.
  • Raad emphasizes the importance of 'taste' and irrational quality investment. He cites building their own terminal framework as an irrational move that became a key differentiator against competitors like Cline.
  • Raad notes that old software patterns like Domain-Driven Design are becoming more useful again because they provide guardrails for 'a bunch of idiots' - AI agents that work 24/7.
  • Raad advises engineers to combine software skill with deep industry expertise. Spending a year in any field makes you more knowledgeable than 99% of people, creating a powerful 'unicorn' combination.
  • OpenCode capitalized on Anthropic's clumsy ban of Claude subscriptions by galvanizing competitors. They secured official OpenAI support the next day, turning a crisis into a strategic win.
Also from this episode: (5)

AI & Tech (5)

  • Raad sees product-market fit as a critical phase where AI can worsen decision-making. He says it's easy to respond to every user request or competitor feature, which results in a Frankenstein product.
  • OpenCode's growth exploded from 650k monthly active users in December 2025 to 2.5 million in January 2026 and was around 6.5 million last month.
  • Raad says GPU supply is bottlenecking even companies of OpenCode's size. Demand is growing exponentially while production is linear, causing a capacity crunch and forcing companies to hoard and pay upfront.
  • OpenCode's business model includes Zen, an inference service that hit a $50 million run rate within five or six months, and enterprise control plane software for managing AI tool usage at scale.
  • Raad criticizes viral predictions like '24-29 year olds are the most valuable asset' as defense mechanisms. He says people confidently assert futures where they are winners to manage anxiety about rapid change.
#Agents#Open Source#Big Tech#Startups
Naval
Naval

Waste Tokens, Save Time • May 27

  • Gumo Roush argues engineering excellence is now about building multiplicative software factories, not delivering individual outputs.
  • Gumo claims 100x or 1000x engineers exist in intellectual domains, citing Satoshi, Notch, Brendan Eich, and John Carmack as examples.
  • Roush dismisses token consumption and lines of code as flawed productivity metrics, likening them to outdated management paradigms.
  • Roush's principle for AI use is to waste tokens to save time, arguing models remain cheaper than human labor regardless of quality.
  • He advocates brute forcing problems by throwing multiple AI models at them iteratively, trusting they will improve with each generation.
  • Naval questions if pure software engineering is dead, suggesting hardware founders gain advantage and model training may be the new software.
  • Max Hodak built substantial software using AI since December, fulfilling long-held project fantasies without writing a single line of code.
  • Hodak states AI removes the intrinsic frustration of debugging, fundamentally changing the learning process for programming.
Also from this episode: (6)

AI & Tech (6)

  • Max Hodak observes AI model performance heavily depends on user judgment and feedback, especially the quality of reprompting.
  • Hodak resisted learning prompt tricks, assuming model improvement would outpace his skill acquisition and preferring a hamfisted approach.
  • Blake Shawl notes AI models now act as principal engineers by proposing architectural trade-offs, demanding more intellectual respect.
  • Shawl highlights architectural choices like database or messaging system selection as areas where human taste and judgment still dominate.
  • Naval predicts the human-as-instruction-giver phase is temporary, foreseeing AI agents directly interfacing with APIs and paying with crypto.
  • Roush argues reusable building blocks are critical for AI agents, citing Mitchell Hashimoto's 'block economy' concept for scalable cooperation.
#Agents#Coding#Startups#Enterprise