Raad says AI execution leaves the what-to-build bottleneck untouched
Thursday, May 28, 2026 · from 6 podcasts
- AI agents accelerate execution but don't solve the harder problem of deciding what to build.
- Firms must restructure around AI-first workflows to survive, as internal coordination now costs more than external action.
- Solo founders use AI as a full executive team, replacing specialist agencies and collapsing learning curves.
6 Sources:
Source Intelligence
- Deep dive into what was said in the episodes
Google is Not a Serious Company • May 28
- Theo argues Google is not a serious company, pointing to a year-plus period of no notable frontier releases from its AI labs since Gemini 1.5 Flash, which he describes as a disaster.
- Theo says Google's 'Omni' model concept of anything-in, anything-out has one real use case: video-to-video generation for tasks like adding fire to a background, which he finds not broadly useful.
- Theo argues most LLM-generated text is never read by humans but is still useful for reasoning and tool calls, while unviewed images and videos have no inherent value, questioning heavy investment in those modalities.
- Theo cites a survey showing Midjourney's user share dropping from 45% to 8%, illustrating the rapid churn in image generation tools as capabilities become commoditized within broader platforms.
- Theo states OpenAI's GPT-4o image generation is now useful for creating UI mockups and dashboards as business assets, not just artistic consumption, marking a shift in the modality's value.
- Theo criticizes Google Cloud's reliability, citing a four-year history of issues and a recent incident where Google's algorithm mistakenly deleted Railway's entire account without human oversight.
- Theo ranks the AI lab hierarchy as OpenAI and Anthropic far ahead, with XAI and Cursor as potential contenders, followed by Chinese labs, and Google in last place due to stagnant trend lines.
- Ben discusses Anthropic's monthly compute spend on SpaceX servers, revealed in the SpaceX IPO filing, as $1.25 billion, which constitutes a majority of Anthropic's estimated $1.5-2 billion monthly revenue.
- Theo describes the Manis-Meta acquisition fallout, where Beijing used a policy to undo the completed $2B deal after employee onboarding, forcing Manis to try raising $1B to buy itself back from Meta.
- Theo contends Google's core failure is bureaucratic fragmentation, contrasting it with OpenAI's model of individual experts moving between teams and Vercel's company-wide 'unblock me' Slack channel that treats internal blocks as P0 issues.
- Ben introduces Lakebed, his integrated cloud framework built in four days with GPT-4o, designed to compile a full-stack app from code with three commands, eliminating the glue work between databases, auth, and hosting.
- Ben argues the pain of deployment has become disproportionate now that AI can build apps in 40 minutes, making the traditional 3-5 hours of cloud configuration feel like an unacceptable bottleneck.
- Ben states Lakebed automatically syncs environment variables from a local .env file to production on deploy, adding, updating, or deleting them as needed, which he calls the right approach for 90% of apps.
Also from this episode: (5)
Models (1)
- Ben reveals a private software engineering benchmark showing GPT-4o and Claude 3.5 Opus leading, with a steep drop to Sonnet 3.6 and Gemini 1.5 Flash, and a final cliff to Gemini 1.0 Pro at 10% performance.
AI & Tech (4)
- Theo argues the Manis case and China's move to close-weight models like Qwen signify a deliberate decoupling from Western AI development, ending the era of Chinese open-weight models feeding the global ecosystem.
- Ben details Cursor's Composer 2.5 training techniques, including reverting implemented features to generate synthetic chat logs for RLHF and using a teacher-student method to correct tool-calling errors without explicit context.
- Ben asserts Google's models fail at reasoning, citing their tendency to get stuck in loops or berate themselves in traces, and posits that adding reasoning was the moment Gemini fell apart competitively.
- Theo describes a novel prompt injection attack vector called 'font hacking', where a PDF uses custom glyphs to show one city name to a human but a different name to an LLM reading the underlying text encoding.
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 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 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.
- 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 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 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: (4)
AI & Tech (4)
- 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 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.
- 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.
- 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.
The Organizational Singularity: AI-Proof Your Company | EP #258 • May 26
- Ismail says companies must architect around intelligence instead of hierarchy, moving from human-centric workflows to AI-native, agentic systems.
- The 'organizational singularity' centers on recursive self-improvement at the workflow level, enabling companies to learn faster than competitors.
- Ismail warns any high-margin business line can be replicated by a small team using tools like OpenClaude in 60-90 days, making incumbent firms vulnerable.
- Companies should build an AI-native digital twin at the edge, migrating workflows like invoice processing to it while leaving the legacy core untouched.
- Ismail estimates a successful transition yields 100x performance improvements per year, and companies can eventually operate with 10-25% of their current workforce.
- Middle management will shrink by 60% as coordination tasks vanish, while the C-suite shifts to dashboard oversight and exception handling.
- Ismail's intelligence stack has six layers: purpose, sensing, interpretation, decision, orchestration, and learning, wrapped by a governance protocol for agent oversight.
- Agent passports with metadata constraints prevent rogue actions, supported by granular rollback, searchable logs, and human review queues.
- Ismail cites Cognitions Labs achieving 73x ARR growth after going fully AI-native, and says the full industry transition will take five to seven years.
- Ismail's rewrite methodology involves backcasting from a future vision, scoring the company on seven dimensions like organizational drag and AI integration.
Also from this episode: (4)
AI & Tech (3)
- Salim Ismail argues Coase's 1937 'The Nature of the Firm' model is obsolete because coordination costs inside a company now exceed external execution costs due to cheap AI agents.
- Ismail notes 44% of Gen Z workers sabotage AI training to protect their jobs, exemplifying the organizational immune system that blocks change.
- The new book 'Organizational Singularity' will be released as a Claude AI skill to stay updated, not as a static text.
Health (1)
- Peter Diamandis highlights Fountain Life's full-body MRI and early cancer detection screening, noting 3.3% of members had undiagnosed cancers.
Behind the Scenes: Using AI to Build a Real Business in Real Time (JWP125) • May 25
- Jake Woodhouse sees AI adoption as more immediate than Bitcoin, noting AI already permeates daily life while understanding Bitcoin requires deeper financial inquiry.
- Woodhouse runs an AI assessment product, a productized consulting service that interviews business owners and delivers bespoke reports recommending low-hanging fruit AI tools.
- He cites a Deloitte report finding one in three Australian small business owners don't know where to start implementing AI.
- Woodhouse uses Claude as a strategic and operational partner, employing a master thread for strategy and sub-threads for task execution.
- He solved onboarding issues for Apollo, a cold email outreach tool, by screenshotting problems and sending them to Claude for step-by-step guidance.
- Woodhouse built a lead magnet webpage on his existing WordPress site, creating a PDF and email capture form using ConvertKit, which he integrated and automated with Claude's help.
- Claude assisted him in updating DNS records on Squarespace to ensure email deliverability, a task he would have previously outsourced.
- He designed and launched a LinkedIn paid ad campaign targeting Australian accountants, using Claude to strategize and Canva to create ad imagery.
- Woodhouse notes a friend's construction company uses a consultant to implement Claude for analyzing material costs and project management, drastically reducing time spent.
- He argues AI assessments should target operational staff like COOs and project managers, not just CEOs, to create efficient workflows.
- Woodhouse claims he built a complete marketing funnel from scratch in one week despite having no technical background and multiple other commitments.
Also from this episode: (2)
AI & Tech (2)
- Woodhouse built a dedicated landing page for the LinkedIn ad traffic on WordPress, again using Claude for copy and design guidance.
- He asserts AI tools enable execution two to four times faster than traditional methods like Google searches, podcasts, or business books.
The AI Daily Brief: Artificial Intelligence News and Analysis
Nathaniel Whittemore
The 4 AI Team Members Execs Should Hire Right Now • May 25
- Nufar Gaspar identifies three common archetypes among executives lagging in AI adoption: the 'podcast CTO' who knows every release but hasn't built a system, the 'weekend tinkerer' who builds privately but not operationally, and the 'manifesto writer' with a vision who hasn't personalized AI use.
- Gaspar argues the leader's quality of AI usage is the single biggest predictor of their team's AI adoption, and leaders who are the best users create the most forward-looking AI organizations.
- Gaspar presents five non-negotiable operating principles for executives using AI: use voice/dictation over typing to capture unstructured thinking, habitually brain dump undocumented context, let AI 'interview' you before complex tasks to surface blind spots, separate planning from execution for critical tasks, and be intentional about where in a workflow your human judgment adds the most value.
- Gaspar advises building a digital workforce with four AI 'team members': a Research Analyst, a Strategic Thought Partner, a Communication Expert, and an Operational Powerhouse, which provide capabilities beyond human bandwidth.
- Before acting on AI research, Gaspar suggests running outputs through three questions: is it grounded in real sources or just AI pattern matching, what's missing that I didn't think to ask, and would you feel comfortable putting your name to it.
- For strategic AI advising, Gaspar recommends building a 'board of advisors' with distinct personas and decision-making styles that debate a decision before presenting it, and calibrating the AI's pushback to match your personal decision-making style.
- For operational AI, Gaspar says leaders should not just automate existing tasks but conceive of dashboards and reports they'd build with unlimited headcount, and they should manually test any new automated brief or process for one to two weeks before committing to full automation.
- Gaspar states the natural progression after mastering the four digital team members is to build an AI 'chief of staff' that orchestrates across them, providing a cross-functional view of decisions and priorities.
- Gaspar emphasizes focusing on the methodology and results of AI systems over specific tool features, advising executives to 'sweat what you're building and how you're building it' rather than the tool choice.
- Gaspar's training is based on working with executives across 30 different countries, observing recurring patterns in how leaders engage with AI.
Also from this episode: (3)
AI & Tech (3)
- For AI research, Gaspar recommends using 'wisdom of the crowd' by running the same query across multiple AI models or sessions, aggregating consensus results, and using a separate model to fact-check the aggregated findings, arguing consensus likely indicates factual accuracy.
- To make an AI communication expert write in your voice, Gaspar advises style profiling by feeding AI your best writing samples for analysis, and creating detailed personas of your target readers to have them review drafts for clarity and impact.
- When giving AI feedback on writing, Gaspar recommends scoring outputs on specific dimensions like clarity and conciseness instead of giving generic critiques, which allows the model to understand precisely how to improve.
The AI paradox: More automation, more humans, more work | Dan Shipper • May 24
- Dan Shipper predicts the 'AI job apocalypse' is not happening, and instead, AI will create more work for humans. He is extremely bullish on product managers and full-stack designers as roles that will thrive.
- Shipper's company, Every, doubled in size from 15 to almost 30 people in one year despite being AI-forward, counter to expectations that AI reduces hiring.
- Work will bifurcate into two modes: everyone will have at least one async agent to delegate to, and most knowledge work will happen inside agent environments like Cursor or Claude Co-work.
- The architecture for workplace agents will shift from personal agents to a 'super agent' per company. Shipper flipped his view after seeing personal agents like OpenClaw require too much maintenance and a human who cares about them.
- Shopify and Ramp are cited as examples of companies already using a single, company-wide agent.
- Shipper predicts the primary work surface will become agent environments with built-in browsers. He uses Cursor's desktop app as his daily driver, running apps like Proof inside its browser so the agent can see and interact with his work.
- This shift changes SaaS economics: users bring their own AI agents and tokens to SaaS tools, potentially saving SaaS companies from paying for AI integration costs themselves.
- Shipper is bullish on SaaS stocks, arguing the 'SaaS apocalypse' narrative is wrong. Agents will increase the number of users of SaaS products, not eliminate them.
- He argues the CLI era for AI work is over, as GUIs within tools like Cursor and Claude Co-work provide a superior experience, especially for non-programmers.
- Automation creates more human work because every agent needs a human to manage, garden, and ensure it works correctly. This 'forward deployed engineer' role is becoming critical.
- Shipper predicts acceptance of AI-generated writing for internal documents and email, valuing the output if the human stands behind the content, not if it's slop they haven't reviewed.
- The key to success is to 'ride the models': be curious, playful, and apply new models to your domain. The edge of AI is wherever it meets a real human doing something, not just in San Francisco.
Also from this episode: (2)
Coding (1)
- Shipper created a 'senior engineer benchmark' testing AI's ability to rewrite sloppy code. GPT-5.5 scored 62/100 using an Opus 4.7 plan, while human senior engineers score in the high 80s.
AI & Tech (1)
- He observes that AI makes 'yesterday's human competence' cheap and commoditized. Human value lies in using that commoditized capability to create something new and interesting.
Related Stories
BUSINESS
SpaceX IPO drains Bitcoin liquidity for orbital AI dominance
Plebchain Radio · BTC Sessions · Bitcoin And | Bitcoin & Economic News · Moonshots with Peter Diamandis · All-In with Chamath, Jason, Sacks & Friedberg · The Intelligence from The Economist
AI & TECH
Manna's 20-cent drones target US suburbs as Pentagon scrambles
This Week in Startups