Software engineers at companies won’t only face competition from vibe-coders, but from their own CEOs too.
Tobias Lütke and Brian Armstrong — the chief executives of Shopify and Coinbase respectively — have both recently showcased GitHub contribution graphs suggesting a dramatic return to hands-on coding. Their activity comes as AI-powered coding agents like Anthropic’s Claude Code and OpenAI’s Codex have lowered the barrier for technically trained executives to re-engage with software development, even after years away from the keyboard.
The Data Speaks For Itself
The evidence is striking. A chart shared of Lütke’s GitHub profile (@tobi) shows contributions that flatlined near zero for much of the past decade, then spiked dramatically to 2,000 in early 2026 — a near-vertical climb that almost certainly correlates with the rise of AI coding tools.
Armstrong posted his own GitHub contribution graph on X (formerly Twitter) on February 25, 2026, with characteristic candor: “Beware, it’s 2026 and the CEO is back in the code base thanks to AI.” His graph, while more modest, shows a clear uptick in recent months.
CEOs Who Were Once Engineers
Neither man is a stranger to code. Armstrong studied Computer Science and Economics at Rice University, earning both a bachelor’s and master’s degree, before going on to build Coinbase into one of the world’s most valuable crypto exchanges. He spent years as a hands-on engineer before the administrative weight of running a publicly traded company inevitably pulled him away from writing software directly.
Lütke’s path into programming is arguably even more foundational. After completing just 10th grade, he enrolled in a German apprenticeship program to learn software engineering and, at 17, was already writing code professionally for a subsidiary of Siemens. That early, visceral love of programming became the seed from which Shopify grew — but scaling a commerce platform serving millions of merchants meant the code gradually gave way to board meetings, fundraises, and org charts.
For both men, the return to coding isn’t a vanity project. It reflects something deeper about what AI agents are making possible for people who once knew how to build but lost the daily practice.
AI as the Great Re-Onboarder
The core problem for technically skilled executives has never been understanding code — it’s been the friction of re-entry. Codebases grow enormous and unfamiliar. Languages and frameworks evolve. The cognitive overhead of context-switching between high-level strategy and low-level implementation is simply too high for most leaders to sustain.
AI coding agents are dismantling that friction. Tools like Claude Code allow developers — and lapsed developers — to interact with a codebase through natural language, ask questions about unfamiliar modules, generate boilerplate, review diffs, and ship small changes without needing to hold the entire architecture in working memory. Codex and similar tools offer comparable capabilities: an always-available pair programmer with encyclopedic knowledge of syntax and patterns, ready to bridge the gap between intention and implementation.
For a CEO who last wrote production code years ago, these tools function as a kind of re-onboarding engine — one that compresses what might otherwise take weeks of catch-up into hours. The executive can focus on what they still know better than anyone: the product intent, the business logic, the user problem. The agent handles the translation into working code.
A Cultural Signal As Much As A Technical One
Lütke has been unusually explicit about his views on AI’s role in software development, previously writing an internal memo emphasizing that Shopify employees should be integrating AI tools into their workflow or risk falling behind. Armstrong has echoed similar sentiments about crypto and technology broadly. Both leaders are clearly walking that talk — and doing so publicly.
There is, of course, a performative dimension to sharing a GitHub contribution graph on social media. But the underlying message being sent to their engineering organizations is real: the CEO is in the codebase. The tools have gotten good enough that the person responsible for product strategy is also willing to be responsible for product implementation, at least at the margins.
That has downstream implications. It shortens feedback loops. It signals cultural expectations about engagement with technical craft. And it raises, in a pointed way, the question of what it now means to be a “technical” leader in the AI era — when the barrier between understanding software and writing it has never been lower.
What This Means For Engineers
Software engineers watching their CEOs post contribution graphs might feel a complicated mix of emotions. On one hand, a CEO who writes code tends to be a more empathetic engineering leader — someone who has felt the pain of a bad abstraction or an underdocumented API. On the other hand, the same AI tools that are helping Lütke and Armstrong return to the codebase are the ones quietly reshaping what it means to write software for a living at every level of the organization.
Vibe coding — the practice of generating functional code through conversational prompting with minimal traditional programming knowledge — has already begun changing who counts as a developer. The CEO contribution graph is a data point in that same trend. If AI can bring a decade-removed executive back into active software development, the implications for how organizations hire, structure engineering teams, and think about technical skill are only beginning to come into focus.
For now, though, the headline is simple: the CEOs are coding again. And they have the commit history to prove it.