Price:

AI & TECH

Agents trigger SAS apocalypse as AI revenue hits $650B

Tuesday, May 5, 2026 · from 6 podcasts, 7 episodes
AI Daily BriefLenny's PodcastThis Week in StartupsPeter McCormack ShowThe a16z ShowSequoia Capital
  • AI agents are replacing per-seat software subscriptions, triggering a $400 billion 'SAS apocalypse'.
  • Enterprise AI spending is real, with cloud giants planning $650 billion in capex this year.
  • The new competitive edge is 'harness engineering,' not model size, as runtimes dictate performance.

The AI bubble narrative is dead, replaced by a grim reality for legacy software. Nathaniel Whittemore describes a 'SAS apocalypse' where autonomous agents like Claude Co-work and Codex are cannibalizing per-seat subscriptions. When an agent can build a bespoke tool for pennies, the middleman software layer becomes a liability. This isn't speculative; hyperscalers are backing the shift with a planned $650 billion in capital expenditure this year.

"Markets stopped asking if AI would work and started fearing it worked too well."

- Nathaniel Whittemore, The AI Daily Brief

The money is moving from traditional enterprise software into the raw compute needed to run the agents that replace it. Google Cloud’s revenue jumped 63%, AWS returned to 28% growth, and Microsoft reported 20 million paid Copilot seats. Amazon is spending every dollar of profit on AI infrastructure, with demand for its Trainium chips so high it would constitute a $50 billion business if booked as revenue.

This shift has moved competition from model weights to 'harness engineering.' The runtime environment - the persistent memory, tool dispatch, and sandboxing - is now the primary location of intelligence. Benchmarks show GPT-5.5’s performance jumps from 61.5% to 87.2% when moved into a superior harness like Cursor’s. Sam Altman recently acknowledged it's increasingly hard to tell where the model ends and the harness begins.

Legacy platforms built for human administrators are uniquely vulnerable. Joe Schmidt IV argues that while systems like Workday have 97% retention, they are 'most important and least loved.' AI-native migration tools can now map and move complex databases in 60 days instead of 12 months, dissolving the friction that protected incumbents. Schmidt contends much reported AI revenue, like Workday’s $400 million ARR, is 'procurement innovation' - flex credits sold on top of old architecture - not true agentic transformation.

"The core user experience of Workday is broken... It took me six and a half minutes to find my compensation data."

- Joe Schmidt IV, The a16z Show

Andrej Karpathy frames the human role in this new stack as directing a fleet of 'intern entities.' The goal is coordinating stochastic agents to maintain a professional bar. As execution is automated, Max Schoening of Notion argues that individual 'agency' - the conviction the world is malleable - becomes the key differentiator. The first 10% of any project is now free; the winners are those who start without permission.

The frontier is autonomous business ownership. Experiments like the Valerie vending machine, run by an OpenClaw agent with a bank account and legal trust, point to a future of zero-employee companies. Startups like Pulsia are already hitting millions in revenue with a single founder orchestrating agents. The capability overhang is widening, and the divide between companies with working agents in production and those stuck in the pilot phase is becoming exponential.

#Agents#Enterprise#Big Tech
Andrej KarpathyNathaniel WhittemoreSam AltmanAmazonClaudeCodexGoogle CloudCursorGoogleGPTMicrosoftOpenClaw

Source Intelligence

- Deep dive into what was said in the episodes

The AI Daily Brief: Artificial Intelligence News and Analysis
The AI Daily Brief: Artificial Intelligence News and Analysis

Nathaniel Whittemore

Is AI Doom Going Out of Style? • May 4

  • Nathaniel Whittemore defines "AI's second moment" as the rise of workable agentic systems, following the first moment of viable AI assistant experiences like ChatGPT. This quarter was deemed the most consequential for AI since ChatGPT launched, with capabilities scaling from 100 million users in 5 weeks to billions weekly.
  • Claude Co-work, launched in January, expanded agentic capabilities to general knowledge work, reportedly triggering emergency meetings at Microsoft. Q1 saw more frontier capabilities shipped than any prior quarter, with the latest Gemini, GPT, and Claude models constantly vying for narrow leads across various benchmarks.
  • Ramp statistics show Anthropic captured 70% of first-time enterprise AI buyers, with OpenAI at 25%, despite OpenAI's higher annualized revenue of around $25 billion. Anthropic hit a $19 billion run rate and rapidly brought Open Claw-like capabilities into its native Claude Code and Claude Co-work ecosystem.
  • Gartner projects that by the end of 2026, 40% of enterprises will have working agents in production, with new tools like agent credit cards from Ramp and Stripe enabling more direct agent spending. The company Pulsia achieved $6 million in annualized revenue with a single founder and zero employees, challenging traditional company design.
  • Monthly pulse surveys show increased AI usage among practitioners, with over 71% having "vibe coded" and 62% using automation or agentic capabilities in the past month. The dominant value derived from AI shifted from time savings (13.6% in February) to increased output and throughput, and new capabilities.
  • AI is creating entirely new functions, such as Generative Engine Optimization (GEO), which helps companies appear more frequently and positively in AI chatbot responses. The GEO market, valued under $1 billion in 2025, is projected to grow to nearly $34 billion by 2034.
  • OpenAI faced backlash and a 775% surge in one-star ChatGPT reviews after signing an agreement with the "Department of War" on the same night as the Anthropic ultimatum, propelling Claude to the top of the App Store. President Trump also secured agreements from hyperscalers to mitigate AI infrastructure costs for Americans.
  • Nathaniel Whittemore predicts the capability overhang - the gap between potential and actual AI value - will widen before it closes, increasing the divide between leading and lagging companies. He argues that focusing on new capabilities rather than just time savings will be more profitable for enterprises.
Also from this episode: (6)

Business (2)

  • The economic stakes have grown from speculative venture bets to a planned $650 billion in capital expenditure this year, signaling a major industry reorientation. This includes a projected $400 billion "SAS apocalypse wipeout" and single funding rounds reaching tens or even hundreds of billions.
  • The narrative around AI markets shifted from a "bubble" in Q4 2025 to concerns about AI being "too good" in Q1 2026, exemplified by public recantings of skepticism from investors like Howard Marks. The "SAS apocalypse" saw widespread carnage among public software companies, with Block cutting 40% of staff.

AI & Tech (4)

  • An inflection point over the holidays, marked by new models like Opus 4.5, GPT 5.2, and improved harness capabilities in Claude Code and Codex, transformed the AI landscape. Claude Code, initially misnamed for its non-coding uses, grew from $1 billion to $2.5 billion in annualized revenue in a few months.
  • Q1 2026 became known as the "quarter of open claw," an open-source project that grew from Claude Bot to GitHub's most starred project ever and was eventually recruited into OpenAI. Jensen Huang, Nvidia CEO, called Open Claw "maybe the most important software release ever."
  • AI politics significantly intensified, notably with the Pentagon's battle with Anthropic over Claude's use in military operations, including a raid against Venezuela's Nicolas Maduro. Defense Secretary Pete Hegseth designated Anthropic a "supply chain risk" after the company refused to comply with Pentagon demands, leading to a lawsuit.
  • The Q2 competitive landscape shifts from model superiority to agent platform usage, with the primary battle being between Claude Code, Codex, and Open Claw. A trend of competitive convergence means "every AI product becomes every other AI product," as platforms expand features into similar spaces.
#Agents#Enterprise#Big Tech#Regulation

Why Agents Make Every Job a Startup • May 3

  • Nathaniel Whittemore cites Google Cloud's 63% year-over-year revenue growth as evidence against an AI bubble, citing similar strong growth from Azure (40%), Meta (33%), and AWS (28%).
  • Google reported a $460 billion cloud backlog, up from $240 billion in Q4, which Joseph Carlson described as 'so crazy it literally looks fake.' CEO Sundar Pichai stated AI is now the cloud's largest tailwind, but near-term growth is limited by compute constraints.
  • Google saw a 40% quarter-over-quarter surge in paid enterprise Gemini customers. Its infrastructure now processes 16 billion tokens per minute, a 60% quarterly increase.
  • Amazon said it added more server capacity than any other company in 2025, with Q1 capital spending at $43.2 billion, a 60% annual jump. Free cash flow fell to $1.2 billion as spending soared.
  • Amazon CEO Andy Jassy claims the company's in-house Trainium chip demand is so high that if booked as revenue, it would constitute a $50 billion ARR business, making it a top-three data center chip player.
  • Microsoft reported 20 million paid seats for Copilot, up from 15 million in January. CEO Satya Nadella noted weekly Copilot engagement now matches that of Outlook.
  • Meta raised its 2025 capital expenditure forecast from $135 billion to $145 billion, primarily due to higher component pricing and data center costs, not new builds.
  • The market punished Meta's aggressive spending despite record quarterly revenue of $56.3 billion (up 33%), sending its stock down 5% overnight, as Jim Cramer noted a lack of justification for its spend.
  • Agent capability growth has shifted through three phases: model weights, context engineering, and now harness engineering, where the runtime environment around the model is the primary source of intelligence.
  • A harness provides agents with persistent memory, reusable skills, sandboxing, and standardized protocols, moving intelligence out of the prompt. Akshay argued this is the fundamental shift enabling reliable agents.
  • Sam Altman told Ben Thompson the harness is inseparable from the model in creating capable agents, making it hard to attribute performance gains to just the model or its runtime.
  • Nathaniel Whittemore defines 'harness as a service' as a new infrastructure category where companies sell pre-built agent runtimes, analogous to AWS selling compute, removing the need to assemble the full stack.
Also from this episode: (3)

AI & Tech (3)

  • Whittemore compares the Open Claw era to hobbyist computing kits, while harness-as-a-service platforms like Cursor SDK represent the shift to pre-built, accessible systems that democratize agent creation.
  • Endor Labs found that switching models to Cursor's harness significantly improved benchmark scores; GPT-5.5's functionality score jumped from 61.5% to 87.2%, and Opus 4.7's rose from 87.2% to 91.1%.
  • Cursor SDK's release has enabled rapid MVP building for agentic products, such as Jack Driscoll's Gmail-integrated coding agent and tools for autonomous bug-catching and IT triage.
#Agents#Big Tech#AI Infrastructure#Models#Coding
Lenny's Podcast
Lenny's Podcast

Why cultivating agency matters more than cultivating skills in the AI era | Max Schoening (Head of Product, Notion) • May 3

  • He predicts a shift away from specialized SaaS towards more general tools like word processors and spreadsheets, but believes the 'as-a-service' maintenance model will persist.
  • Schoening says great products have one exceptionally good tiny core, like GitHub's pull request or Heroku's 'git push'. Adding more features to make a product great never works.
  • He warns that merging product roles risks losing specialists needed for high-quality engineering at scale and for deliberate design craft.
  • Schoening uses the Jobs to be Done framework to force teams to zoom out and honestly assess if a user would hire their product, cutting through internal corporate biases.
Also from this episode: (8)

Startups (2)

  • Max Schoening says AI makes the first 10% of any project free, drastically lowering the effort to build a startup's first version.
  • Schoening encourages building agency by making things and tinkering, which reveals that the world is built by people no smarter than you.

AI & Tech (3)

  • Schoening argues the most important trait for product builders in the AI era is agency, not specific skills. He defines agency as the belief the world is malleable and can be changed.
  • He advocates for malleable software, where users have ownership over their computing life and can tweak tools, contrasting it with rigid apps designed by corporate ivory towers.
  • Schoening sees software engineering capabilities improving exponentially but is unimpressed with AI progress in other domains like writing. He believes software will eat the world faster, with coding principles applied everywhere.

Psychology (1)

  • He defines taste as the ability to run a virtual machine in your head to predict if a specific in-group will like an idea. He says you build taste through reps and exposure to other people's work.

Business (2)

  • Schoening's hot take is that universal basic income already exists in the form of knowledge work, arguing we've invented necessary jobs far beyond basic living needs.
  • His contrarian opinion is that inclusivity isn't always great; he believes in small group theory and building exceptional products for a top-tier in-group, even if it means excluding others.
#Agents#Startups#Enterprise#Reasoning
This Week in Startups
This Week in Startups

Can an AI Agent Legally Own a Company? Christian van der Henst's Wild Experiment| E2283 • May 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.
Also from this episode: (10)

AI & Tech (8)

  • Christian van der Henst's team built a vending machine called Valerie, registered as a business owned and operated by an OpenClaw AI agent with access to bank accounts, using Brevan Love to package the agent's IP into a legal trust structure.
  • The Valerie agent autonomously managed a vending business, handling tasks like inventory procurement, dynamic pricing, and market research by scraping sites like Instacart, but faced friction with online payments flagged as bot activity.
  • Jason Calacanis argues the legal and regulatory framework for agent-owned businesses is undeveloped, as KYC processes require human identification and food vending permits remain complex, limiting such experiments to private venues.
  • 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 describes Bit Tensor as an incubator with 128 competitive subnet slots that use the TAO currency, where underperforming projects face relegation to ensure network quality and urgency.
  • Alex Wilhelm notes Anthropic's potential $900 billion valuation creates a Polymarket bet on whether it will flip Bitcoin's $1.58 trillion market cap by year-end, with current odds at 43%.
  • 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.

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.
#Agents#AI Infrastructure#Startups#BTC Markets#Macro
The Peter McCormack Show
The Peter McCormack Show

Peter McCormack

#171 - Michael Saylor - You Have 10 Years Before You're Locked Out Forever • Apr 30

Also from this episode: (10)

Protocol (8)

  • Michael Saylor argues the dollar supply expands at roughly 7% annually, a trend consistent for the last 100 years. This currency debasement transfers wealth from those who don't own appreciating assets.
  • Saylor categorizes global currencies into tiers. The US dollar is the first-tier reserve currency. Second-tier currencies like the pound and euro stagnate against it, while third and fourth-tier currencies collapse.
  • Saylor claims scarce desirable property is the only reliable defense against monetary debasement. Historical examples include prime land, gold, fine art, and intellectual property like music rights.
  • He defines Bitcoin as digital capital and the highest form of capital yet discovered. Saylor argues it perfected portable property rights, allowing global peer-to-peer settlement without intermediaries.
  • Saylor warns that grandiose ambitions break systems, citing Napoleon and the fall of Rome. He applies this to Bitcoin, stating its primary ambition should be to not break the network by adding features.
  • He states that after the block size wars, Bitcoin transaction fees are minimal. Saylor cites Clark Moody's dashboard showing an average fee of $0.32, proving increased bandwidth was unnecessary.
  • Saylor explains his company's digital credit instrument, STRC. It aims to offer ~11.5% yield by stripping volatility from Bitcoin, targeting a market 100x larger than direct Bitcoin investors.
  • He cites a 39% annualized return for Bitcoin over the past five years, double the S&P 500, but notes its high volatility deters widespread adoption for savings.

AI & Tech (1)

  • He posits that human capital is being demonetized by AI and automation. White-collar bots are already here, and physical robots will arrive within a decade, automating both intellectual and manual labor.

History (1)

  • Saylor references historical examples from Murray Rothbard's 'Conceived in Liberty', detailing how American colonies debased currencies and defaulted on debts, leading to Shays' Rebellion and the Constitution.
#Robotics#Macro#BTC Markets#Agents#Banking
The a16z Show
The a16z Show

Workday’s Last Workday? AI and the Future of Enterprise Software • Apr 30

  • Joe Schmidt argues the core user experience of Workday is broken, citing his own six-and-a-half-minute struggle to find compensation data as evidence that no employee enjoys interacting with the portal.
  • Workday's 97% gross dollar retention rate demonstrates the extreme difficulty of displacing entrenched enterprise systems, a defensibility built during the last major platform shift from on-premise to cloud.
  • Schmidt contends that current enterprise AI revenue metrics, like Workday's $400 million AI ARR, are often procurement innovations rather than fundamental product shifts, lacking true agentic experiences.
  • The new platform shift enabling disruption is AI-native architecture, which for the first time allows founders to promise CHROs and CIOs a fundamentally different core system that changes how work is done.
  • An AI-native competitor must enable deployment in 30 to 60 days, a drastic reduction from the historical 12-plus month implementations that required expensive consultants.
  • Schmidt identifies six critical properties for an AI-native Workday successor: rapid deployment, workbench-native customization, agent-first interaction, open APIs, enterprise-grade security, and global compliance readiness.
  • The disruption opportunity is in brownfield replacement, not greenfield sales, as enterprises now have kinetic energy to rip and replace systems where employees are effectively hostages.
  • HR software may become the beacon for mass AI adoption in the enterprise, as its transformation will signal when AI moves beyond early adopters in major cities to broader organizational takeoff.
  • Agent-first HR systems will be critical for permissioning and identity management as more AI agents perform work on behalf of humans, a growing concern for CIOs.
  • Incumbents like Workday are actively fighting the shift, evidenced by executive comebacks, layoffs, and acquisitions like Hired to fend off new competitors.
#Enterprise#Agents#Markets#Big Tech
Sequoia Capital
Sequoia Capital

Sequoia Capital

Andrej Karpathy: From Vibe Coding to Agentic Engineering • Apr 29

  • Karpathy states that OpenClaw's installation exemplifies software 3.0. Instead of a complex bash script, you copy-paste instructions for an agent, which uses its intelligence to adapt to the environment and debug issues.
  • Karpathy argues LLMs enable new applications, like automated knowledge base creation from documents, which couldn't exist before because there was no code to reframe unstructured data.
  • Karpathy distinguishes vibe coding, which raises the floor for all programmers, from agentic engineering, which preserves professional software quality standards while using agents to accelerate development.
  • Karpathy suggests hiring for agentic engineering should involve a large, practical project like building a secure Twitter clone and then stress-testing it with adversarial agents, not puzzle-solving.
  • Karpathy argues that as agents handle more implementation, human skills like aesthetic judgment, taste, system design, and oversight become more valuable, not less.
Also from this episode: (8)

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.
#Agents#Models#Coding#Enterprise

Related Stories