AcceleratingBiz®

Growth CEO’s AI Briefing · Edition #1

Open Season

Saturday, April 4, 2026 · 3 Developments · 2 Perspectives · 2 Case Studies · 1 Resource

Designed by AcceleratingBiz®. Curated, synthesized, and produced by Claude agents. AI-assisted production means occasional errors may appear.

This was a week where several threads came together. The infrastructure for AI-native companies is now available at every scale, from edge devices to data centers, and the first real-world proof points are starting to show up in actual financials. Google released Gemma 4 under Apache 2.0, giving growth companies an open model family worth taking seriously. A solo founder scaled a telehealth startup to $1.8B in projected sales using AI tools and outsourced partners. And Simon Willison, one of the most credible voices in software engineering, made the case that AI coding agents crossed a meaningful threshold in late 2025; the implications for your engineering org are worth thinking through carefully.

The thread connecting these stories is structural: the companies that treat AI as a line item will be outmaneuvered by those that treat it as the operating system.

Developments

Technology

April 2: Google Releases Gemma 4 — Four Open Models, Every Device, Apache 2.0

On April 2, Google DeepMind released Gemma 4, a family of four open-weight models spanning from a 2B-parameter edge variant that runs on a Raspberry Pi in under 1.5GB of memory to a 31B dense model ranked #3 among all open models. All four handle code generation, vision, and multi-step agent tasks. The smallest variants add native audio input and run entirely offline. The Apache 2.0 license is a first for the Gemma line, removing the commercial restrictions that previously pushed enterprise buyers toward Qwen and Mistral alternatives. Developers have downloaded Gemma models over 400 million times since the first generation launched.

So what

For growth company CTOs evaluating their AI stack, Gemma 4 changes the build-vs-buy math. Native function calling and structured JSON output mean you can run sophisticated AI agents locally, whether on customer devices, in air-gapped environments, or wherever data sovereignty matters, without per-token API costs. The 26B MoE variant offers strong performance at a fraction of the compute of full-size models. If your product roadmap includes on-device intelligence or offline agent capabilities, start prototyping this quarter. The Apache 2.0 licensing removes the last legal ambiguity that kept open models out of many enterprise procurement conversations.

Confidence: Established fact** — first-party announcement from Google DeepMind with day-one availability on Hugging Face, Kaggle, Ollama, and Google AI Studio.

Go deeper Google DeepMind Blog · The Neuron

Technology

April 2: Cursor 3.0 Rebuilds Around Agent Orchestration

On April 2, Cursor released version 3.0, a ground-up rebuild centered on agent orchestration rather than code editing. The default view now opens on an agent-centered workspace, with a chat-driven orchestration panel replacing the traditional file tree. A new cloud feature lets an agent build a feature asynchronously and sends the developer a screencast demo when complete. Every's Vibe Check team spent a week testing it and found the orchestration layer impressive but premature. The filesystem sidebar has been demoted, the skills ecosystem is fragmented across Claude Code, Codex, and Cursor with no interoperability, and the core question of "who is this for?" remains unresolved.

So what

Cursor 3.0 matters less as a product review and more as a market signal. The AI-native IDE category is pivoting from "help me write code" to "manage my fleet of coding agents." This is the same shift happening across enterprise software more broadly: tools are moving from augmenting human work to orchestrating autonomous work. If your engineering team uses Cursor, schedule a hands-on evaluation of 3.0. If you're a growth software company, pay attention to what this reveals about your own product trajectory. Your customers may soon expect your tools to manage agents, not just serve humans.

Confidence: Established fact** — first-party release from Cursor; independent testing by Every's product team.

Go deeper Cursor Blog · Every / Vibe Check

Technology

April 2: OpenAI Acquires TBPN — First Media Deal Ahead of IPO

OpenAI announced the acquisition of TBPN, the daily live tech talk show pulling ~70K viewers per episode, in a deal reportedly worth in the low hundreds of millions. TBPN's 11-person team will report to OpenAI's chief of global affairs, Chris Lehane. The show will drop its advertising business but retain editorial independence. Co-founders Jordi Hays and John Coogan launched the live show 17 months ago, and the company was reportedly on pace for $30M in revenue this year.

So what

This is worth noting less for the media angle and more for what it signals about OpenAI's IPO positioning. Buying a direct channel to the tech CEO audience is a reputation management play in a year where OpenAI's public perception has taken significant hits. For growth CEOs, the meta-lesson is that owned media is becoming a strategic asset for AI companies preparing for public scrutiny. The editorial independence claim deserves skepticism given that a chief political operative is overseeing it.

Hype check

"Editorial independence" under a chief of global affairs is a contradiction worth flagging. TBPN has historically been enthusiastically pro-tech; this acquisition is unlikely to introduce adversarial coverage.

Confidence: Established fact** — confirmed by OpenAI's official announcement and TBPN co-founders.

Go deeper OpenAI · The Rundown AI

Perspectives

Business

Simon Willison: "We've Passed the Inflection Point. Dark Factories Are Coming."

Simon Willison, co-creator of Django and one of the most trusted independent voices on AI's impact on builders, sat down with Lenny Rachitsky for a wide-ranging conversation on AI and software engineering. His key arguments are worth considering carefully. He believes November 2025 was the inflection point when AI coding agents crossed from "mostly works" to "actually works," driven by GPT-5.2 and Opus 4.5. He now writes 95% of his code from his phone and is mentally exhausted by 11 a.m. He argues that mid-career engineers, not juniors, are most at risk right now. The next leap is what he calls the "dark factory" pattern, where nobody writes or reviews code and AI handles its own QA. And he warns that prompt injection remains a fundamental, unsolved security problem; the combination of private data, untrusted content, and external communication will likely lead to an AI Challenger-style disaster.

So what

If Simon Willison is saying mid-career engineers are most at risk and dark factories are coming, your CTO needs to hear it directly. The three agentic engineering patterns he describes (red/green TDD, templates, and hoarding) are immediately actionable for any engineering team. The prompt injection warning is especially relevant for growth companies building AI-powered products: if your agents handle private data and communicate externally, you have an unsolved security problem that no amount of prompt engineering will fix. Share this podcast episode with your technical leadership team.

Confidence: Informed perspective** — Simon Willison is among the most credible independent voices in the AI-and-software intersection. His views are experience-based and widely corroborated.

Go deeper Lenny's Newsletter

Business

Agent-First Pricing: Per-Seat Models Break When the User Doesn't Have a Pulse

TLDR Founders highlighted a growing consensus that per-seat SaaS pricing breaks down when the primary "user" of your software is an autonomous AI agent. The shift toward charging for outcomes delivered rather than access provided is already underway, and founders who don't rethink pricing for agent-era customers will be undercut by those who do. Separately, the same edition noted RevenueCat grew developer signups 40% in March alone, driven by the vibe coding wave. That's an early signal of how AI tailwinds create measurable platform effects for companies positioned correctly.

So what

This is directly relevant to virtually every growth-stage B2B software company. If agents are increasingly the ones using your product, pulling data, triggering workflows, and making API calls, then per-seat pricing is leaving money on the table and misaligning value capture. Start modeling what consumption-based or outcome-based pricing would look like for your product. The RevenueCat data point is equally important: companies that position themselves in the path of an AI-driven behavior shift can see 40% monthly growth with no change to the underlying product. What's your equivalent tailwind?

Confidence: Informed perspective** — reflects a growing industry consensus, but specific pricing models remain experimental.

Go deeper TLDR Founders

Case Studies

AI Strategy & Transformation

Medvi: How a Solo Founder Used $20K and AI Tools to Build a $1.8B Telehealth Company

The New York Times reported on April 2 that Matthew Gallagher launched Medvi, a GLP-1 telehealth startup, from his living room in September 2024 with $20K and a stack of AI tools. He used ChatGPT, Claude, and Grok for code and operations; Midjourney and Runway for ad creatives; and ElevenLabs plus custom AI agents for customer service. He outsourced the regulated clinical and pharmacy infrastructure to CareValidate and OpenLoop. In its first full year, Medvi generated $401M in sales, served 250K customers, and posted a 16.2% net profit margin, roughly 3x the margin of publicly traded competitor Hims & Hers with 2,400 employees. Medvi is now on pace for $1.8B in 2026 revenue with a total headcount of two: Gallagher and his brother.

So what

Before your board starts forwarding this as evidence that AI replaces entire companies, apply the hype filter. Medvi chose a market with desperate consumer demand (GLP-1 drugs), outsourced all regulated complexity, and executed extremely well on direct-to-consumer marketing. The transferable lesson isn't "fire everyone." It's that AI tools now compress the traditional startup playbook so dramatically that a well-positioned solo operator can outmaneuver incumbents on speed, margin, and customer acquisition cost. For growth CEOs, the question is: where in your business could a team of two, armed with the right AI tools and outsourced partners, achieve what currently takes twenty?

Hype check

The FDA issued warning letters to dozens of telehealth companies including Medvi in March 2026 regarding compounded GLP-1 drug safety. The AI customer service chatbot fabricated drug prices (which Gallagher honored) and hallucinated products that didn't exist. The single-point-of-failure structure, with Gallagher personally answering 1,000+ customer service calls, is fragile. The NYT verified financials, but 2026 revenue projections are the company's own.

Confidence: Informed perspective** — NYT-verified financials establish the revenue claims; the sustainability and replicability of the model remain unproven.

Go deeper NYT · The Rundown AI

People & Organization

SIDEARM Sports: The President Went First on AI. Everyone Else Followed.

Section AI's AI:ROI newsletter profiled how SIDEARM Sports president George Scott drove company-wide AI adoption by getting deeply fluent in AI himself first. He took courses, did live tool demos at company All Hands meetings, and made AI proficiency part of every employee's performance review. The company selected a curated tool suite (Claude, Gamma, Perplexity), required a defined proficiency baseline before anyone could join the AI committee, and built accountability into meetings, dashboards, and governance structures. VP of Operations Elizabeth Pietrzak led the operational rollout, building the benchmarks, governance, and AI committee structure that made adoption systemic rather than aspirational.

So what

This is one of the most practical AI adoption case studies for growth company CEOs because it addresses the hardest problem: getting from "we have an AI strategy" to "every employee is AI-proficient." Three transferable lessons stand out. First, the CEO or president must be the first "head of AI"; George Scott demoed tools live before asking anyone else to use them. Second, don't let enthusiasm substitute for capability. Require a real baseline before building governance structures. Third, frame AI proficiency as a career imperative for each employee, not just a company initiative. If you're struggling with AI adoption across your organization, this is the playbook to follow.

Confidence: Established fact** — first-party case study from Section AI based on direct interviews with SIDEARM leadership.

Go deeper Section AI (AI:ROI)

Resources

Learning

Anthropic's $1B-to-$19B Growth Run: What Growth Leaders Can Learn from Claude's Scaling Playbook

Lenny Rachitsky interviewed Amol Avasare, Anthropic's Head of Growth, on scaling from $1B to $19B ARR in 14 months. Several takeaways are worth flagging for growth leaders. Anthropic indexes 70/30 toward big bets, which is the opposite of most growth teams. Activation is the single highest-leverage growth problem in AI products. The company uses an internal tool called "CASH" to automate growth experiments with Claude. Avasare uses Cowork to detect team misalignment in Slack. Their focus on AI coding created a research flywheel that accelerated their own models. And the ratio of PMs to engineers may need to flip as AI makes engineers exponentially more productive.

So what

Whether or not you're selling an AI product, Anthropic's growth playbook is worth studying because it represents the fastest B2B scaling in history. The insight that activation, not acquisition, is the highest-leverage growth problem in AI products applies directly to any growth company adding AI features. If users try your AI capability and don't immediately see value, they don't come back. The idea that PM-to-engineer ratios may need to invert as AI coding accelerates should be on every growth CTO's radar as they plan 2027 headcount.

Confidence: Informed perspective** — first-party account from Anthropic's Head of Growth on Lenny's Podcast. Revenue figures are Anthropic's own claims.

Go deeper Lenny's Newsletter

Sources in this edition: The Neuron, The Rundown AI, Lenny's Newsletter, Every (AI & I / Vibe Check), Section AI (AI:ROI), TLDR AI, TLDR Founders, AINews (swyx), The Deep View, Superhuman AI, AI Mindset, TheSequence, Not Boring, Turing Post, AI Scale Tips, Daily AI, Latent Space, The New York Times, Fortune, PYMNTS, The Economist

← Latest Editions and Search