Price:

AI & TECH

Agents slash software costs to zero as startups face vertical integration

Monday, May 11, 2026 · from 2 podcasts
  • Claude Code's $2.5B run rate proves AI agents can build and maintain software for near-zero cost.
  • Startups like Cursor race to build middleware before smarter models make them obsolete.
  • Block cut 40% of staff, signaling a broader SaaS revenue collapse driven by AI efficiency.

The software industry's economics are breaking. Six months after new models like Opus 4.5 and GPT 5.2 launched, the cost of building software is approaching zero. On The AI Daily Brief, Nathaniel Whittemore points to Anthropic’s Claude Co-work, a platform built entirely by Claude’s own coding agent, as the proof point. It triggered emergency meetings at Microsoft.

Revenue figures show the scale. Claude Code hit a $2.5 billion annualized run rate in just two months. Cursor doubled to $2 billion. This isn't speculative venture capital - it’s $650 billion in corporate capex for 2026 alone being redirected. Whittemore describes a "SAS apocalypse" where traditional enterprise valuations crater as AI-native tools cannibalize their functions.

"When Anthropic released Claude Co-work, it triggered emergency meetings at Microsoft. The platform was built entirely using Claude’s own coding agent, proving that the cost of building and maintaining software is approaching zero."

- Nathaniel Whittemore, The AI Daily Brief

The pressure is forcing a brutal efficiency race. On Nerd Snipe, Ben observes that incumbents like Microsoft and Google struggle to adapt as the lead time for new AI features shrinks to months. Startups like Cursor are in "internal fire" mode, desperately building the scaffolding - or "harness" - that makes raw models useful for development.

Their moat is shallow. Ben argues that if a model gets smart enough to handle its own context and tools, the complex middleware becomes technical debt. OpenAI is already lapping teams of 1,500 engineers with a fraction of the headcount. For now, the "folding table boys" who move fast are winning.

"The future belongs to the leanest harness. If a model gets smart enough to handle its own context and tools, the complex middleware startups are building today becomes technical debt."

- Ben, Nerd Snipe

The human impact is immediate. Companies like Block are cutting 40% of their staff. The market logic has flipped: investors now fear AI is too effective for the old software business to survive. The gap is widening between companies using agents to automate tasks and those, like the $6M-revenue firm Pulsia, using them to create agentic companies with one founder and zero employees.

Source Intelligence

- Deep dive into what was said in the episodes

The New Jobs AI Will CreateMay 10

  • Nathaniel Whittemore identifies Q2 2026 as AI's "second moment," marking a shift from viable assistant chatbots to workable agentic systems, which he deems the most consequential period since ChatGPT's launch.
  • Nathaniel Whittemore points to Q4 2025/Q1 2026 as an inflection point, driven by new models like Opus 4.5 and GPT 5.2, alongside transformative capabilities from Claude Code and Codex, leading to record frontier model releases.
  • Q1 2026 was the "quarter of Open Claw," an agent project that evolved from Claude bot, became GitHub's most starred open-source project, and was subsequently recruited into OpenAI. Nvidia's Jensen Huang deemed it a highly significant software release.
  • Enterprise AI adoption is shifting from pilots to production agents; Gartner forecasts 40% of enterprises will have working agents by year-end 2026, supported by new financial tools for agents to spend money.
  • Q1 2026 saw a "SAS apocalypse" as investor concern shifted from AI's potential insufficiency to its disruptive power, evidenced by Block cutting 40% of its staff, alongside substantial AI company revenue growth.
  • Leading AI companies reported immense revenue growth, with Claude Code reaching $2.5 billion annualized revenue, Cursor doubling to $2 billion, and Anthropic achieving a $19 billion run rate.
  • Practitioner surveys show widespread AI usage, with 71% engaging in "vibe coding" and average users employing 3.5 models. The perceived value of AI shifted from time savings to increased output and new capabilities.
  • AI adoption varied across sectors: HR deployments grew 320% in 12 months, and sales emerged as the most mature function with 63% of use cases "primetime." Finance adoption was high, but 91% reported low impact due to data quality.
  • Marketing is creating new fields like Generative Engine Optimization (GEO), projected to grow from under $1 billion in 2025 to $34 billion by 2034, as user behavior shifts towards chatbot-based search.
Also from this episode: (5)

AI & Tech (4)

  • The "AI second moment" signifies dramatically scaled capabilities, with weekly active users reaching billions and economic stakes involving $650 billion in projected capital expenditure this year, signaling a major corporate reorientation.
  • Nathaniel Whittemore describes Anthropic and OpenAI converging in strategy; Anthropic gained 70% of first-time enterprise AI buyers, even as OpenAI, with higher overall annualized revenue of $25 billion, sought to consolidate its products.
  • AI politics escalated significantly, highlighted by the Pentagon's dispute with Anthropic over military AI use, resulting in Anthropic's designation as a supply chain risk, an unprecedented move for a US company.
  • OpenAI's agreement with the Department of War triggered a 775% surge in one-star reviews for ChatGPT, simultaneously propelling Claude to the top of the App Store and demonstrating public sensitivity to AI ethics.

Labor (1)

  • Nathaniel Whittemore notes widespread societal destabilization, with job exposure fears and rumors of 20% Meta layoffs contrasting with advancements like AI-designed cancer vaccines and "zero employee" companies like Pulsia ($6M annualized revenue).

Theo Almost Lost $1 MillionMay 6

  • Theo discovered Azure's caching was broken, with a 0% cache hit rate, because a 'noisy neighbor' caused their cache implementation to buckle.
  • GPT-5.5's latency was 26 seconds in a benchmark, beating Grok 4.3 (1 minute), GPT-4-mini (32s), Claude Haiku 4.5 (35s), and Codex Spark (39s), despite having the lowest tokens-per-second.
  • GPT-5.5 averaged 8 tool calls in the benchmark, while Grok 4.3 did 19, GPT-4-mini did 15, Claude Haiku 4.5 did 11, and Codex Spark did 23.
  • Theo argues smarter models like GPT-5.5 are cheaper in practice because they use fewer tool calls and tokens, but Ben counters that tool call frequency does not always correlate with intelligence.
  • Ben uses OpenClaw on a dedicated phone to run automated tasks like data pipeline monitoring and email inbox triage, but restricts it to read-only access for safety.
  • The OpenClaw repository contains 2.6 million lines of TypeScript, which Ben calls the 'ultimate slop factory,' compared to T3 Code's 200,000 lines.
  • Running the 'GStack' skill in Conductor consumed 37% of a $20/month Claude subscription in one session, creating an empty repo on Ben's GitHub.
  • Ben now defends Gary Tan's 'GStack' concept after using the 'Impeccable' skill, which injects a live design editing UI into a dev server, a pattern he had proposed in an email a year prior.
  • Ben argues the 'Garry's List' and lines-of-code maxims are useful for shifting the Overton window, pushing developers to use code more creatively to explore ideas rather than treating it as expensive.
  • The primary coding agent SDKs are Cursor, Open Code, Claude, Codex, and Pi, with Pi being Ben's favorite and Codex offering effectively unlimited inference on its subscription.
  • BAML is a structured output SDK that unfixes JSON formatting from models like GPT-4, which Ben used for a project with complex object shapes.
Also from this episode: (2)

AI Infrastructure (2)

  • Theo posted a benchmark showing Azure's OpenAI inference was 4-10x slower than OpenAI's direct endpoints, and a Microsoft executive responded to the public criticism to fix the problem.
  • After the fix, Azure's OpenAI inference became 10-20% faster than OpenAI's direct endpoints, allowing Theo to finally use his $1M Azure credit.