Silicon Valley GirlGitHub CEO: Why Now Is the BEST Time to Be a Developer | Thomas Dohmke
CHAPTERS
AI won’t eliminate the need for developers (teaser & setup)
Marina introduces Thomas Dohmke and frames the central tension: AI coding tools are booming while big tech hiring slows. Thomas immediately pushes back on the idea that you can build a durable billion-dollar company without real development skills.
- •Who Thomas is (GitHub CEO; Copilot’s growth; Microsoft integration context)
- •Question: will developers still be needed soon?
- •Claim: “AI + no coding skills” as a path to huge business is a misconception
- •Premise: if everyone can do it instantly, differentiation disappears
What “vibe coding” actually means in practice
Thomas defines vibe coding as working inside an IDE with an AI agent in “agent mode,” where the user largely follows along rather than writing and reviewing every line. The focus shifts from code-level intent to task-level interaction with the agent.
- •Vibe coding = delegating tasks to an AI agent inside tools like Copilot/Cursor
- •User interacts with suggestions and runs commands more than writing code
- •Less continuous code review; more conversational steering
- •Agent-first workflow changes how beginners and non-coders build
How far vibe coding can take you—and where it breaks
You can build surprisingly functional apps (auth, settings, basic web apps), but complexity eventually forces you to understand the code. Thomas compares it to prompting image generators: iterative prompting works until you hit a wall, then you need manual editing skills.
- •Capability scales with your patience and prompting skill
- •Prompting becomes a ‘game’ of trial-and-error without code understanding
- •Comparison to image prompting workflows (iterate until stuck)
- •Limits appear with performance, scaling, and deep complexity (e.g., Black Friday scale)
- •At the edge, you must read/modify source code like a pro developer
“No developers in 2 years?”—why businesses still need real engineering
Marina asks if vibe coding could replace developers for building major companies. Thomas argues that if AI enables anyone to build the same thing quickly, the product becomes commoditized; competitive advantage shifts to teams that build more complex, differentiated systems.
- •To compete in tech, you still need development capability
- •If everyone can build it in minutes, SaaS value and moats shrink
- •AI enables startups to build 10x–100x more complex products
- •Differentiation comes from sophistication, not just speed-to-prototype
- •Not all businesses require coding, but tech businesses do
Prompting as leverage (sponsor segment)
Marina pivots to the idea of AI as a “co-founder” and emphasizes that prompting skill determines output quality. She highlights a HubSpot-made prompt engineering guide and frameworks for structuring prompts and building reusable prompt components.
- •AI can act like a co-founder if you communicate well
- •Better prompting = more control and better results
- •ROSES framework: role, objective, scenario, expected output, steps
- •Modular prompts as reusable components to save time
- •Positioning: systems over ‘random prompting’
Why there will be more developers, not fewer
Thomas predicts a surge in developers because AI lowers the barrier to entry and helps people get unstuck while learning. He also distinguishes between “consumer developers” building personal micro-apps and professionals building complex systems and agents.
- •AI democratizes learning: guidance, debugging help, faster iteration
- •Kids/teens can start by building games and learning by doing
- •AI reduces the ‘stuck with no help’ problem for beginners
- •Two-tier future: consumer micro-app builders vs professional engineers
- •Professional development remains essential for building core AI systems
The smartest companies will hire more—AI as acceleration, not cost-cutting
Thomas argues that multiplying developer productivity changes company strategy: if one developer becomes 10x, scaling teams can yield outsized output. He frames AI as a way to accelerate roadmaps rather than simply reduce headcount.
- •If you 10x a developer, 10 developers can do ~100x output
- •Analogy: websites became easy, but design/ownership and differentiation still mattered
- •Building “web pages for small businesses” is no longer a moat
- •AI expands what teams can attempt, not just what they can save
- •Winning companies invest to move faster than competitors
“Who’s buying?” and the backlog paradox
Marina raises skepticism that output can grow if demand is fixed. Thomas responds that it’s a temporary uncertainty phase; in reality AI tends to create more backlog and more ambition, not less work, as new capabilities generate new requirements and products.
- •Short-term: market uncertainty makes companies cautious
- •AI doesn’t eliminate work; it increases backlogs and expectations
- •No one is ‘finishing the backlog’—AI expands what’s possible
- •Hiring rebounds when companies realize leverage compounds
- •Long-term: more capability drives more demand for software
90% of code by agents—why developers still stay busy
Thomas predicts agents will write the majority of code, but total code volume will grow dramatically. In his math, even if developers write a smaller share, the overall pie expands, leaving substantial human work for orchestration and direction.
- •Prediction: ~90% of code written by agents
- •Interpretation: share decreases, but total volume increases ~10x
- •Developer work remains comparable (or grows) in absolute terms
- •AI’s value is acceleration rather than replacement
- •Humans remain responsible for steering, reviewing, and outcomes
Big tech hiring slowdowns: uncertainty, transitions, and AI mandates
They discuss why companies pause hiring and require teams to “use AI first.” Thomas attributes it to macro uncertainty and a transition where organizations enforce AI adoption, sometimes reshaping teams to match faster operating tempos.
- •Hiring pauses reflect political/economic/tech uncertainty
- •Companies slow down to recalibrate operating models
- •Resistance to AI becomes a competitiveness risk
- •Transition phase: mandates, new expectations, different roles
- •Example of big bets signaling direction (e.g., major AI investments)
Advice for learners: adopt AI early—teens have an edge
Thomas’s guidance to new coders is to learn with AI tools from day one. Younger learners can move faster because they have more time, fewer entrenched habits, and greater openness—similar to how Gen Z grew up native to smartphones.
- •Core advice: learn coding alongside AI agents
- •Young people have time and flexibility to experiment and learn
- •AI-native generation will treat agents as normal tools everywhere
- •Agents will be used beyond coding (email, planning, work workflows)
- •Career advantage: be fluent in the new default toolchain
Best places to start vibe coding (tools and where people get stuck)
Thomas says it’s the right time to try vibe coding because multiple ecosystems now support it, from chat-first tools to deployable app builders. He notes the common failure modes: not knowing what to ask, shallow prompts, and inability to modify source code when edge cases arise.
- •Chat/agent options: ChatGPT/Claude; OpenAI Codex mentioned
- •App builders and deployment-friendly tools (e.g., Vercel-like workflows)
- •“No technical background” tools exist, but users still hit walls
- •Common issues: unclear requirements, insufficient prompting depth, weak UI/quality
- •To progress past the edge: learn source code basics and editing
AI, ideas, and AGI: capability vs creativity and emotion
Thomas believes AI can help humans generate better ideas through reflection, recombination, and reasoning, but he doubts current systems are truly creative. On AGI timelines, he argues it depends on definitions—models may exceed humans in knowledge and summarization, yet lack emotion and sentience that may underpin creativity.
- •AI as a thinking partner: “what am I missing?” and recombination prompts
- •Humans supply the spark; AI expands exploration and packaging (e.g., pitch decks)
- •AGI depends on definition: knowledge/reasoning vs sentience/creativity
- •Claim: current systems aren’t creative; emotion may be tied to creativity
- •Examples that ‘feel like AGI’ (self-driving, vibe coding) but aren’t fully humanlike
Fear, parenting, and staying optimistic: skills that matter + top AI tools
Thomas frames the era as uniquely empowering—anyone can build from anywhere with internet access and AI assistance—and encourages curiosity and problem-solving. He advises people worried about job loss to adopt AI and become the orchestrator, then closes with his top tools: Copilot, ChatGPT, and transcription/summary apps.
- •Parenting lens: teach curiosity, exploration, and independent problem-solving
- •Why now is exciting: turning ideas into apps faster than ever
- •Response to job fear: upskill by mastering AI tools and orchestration
- •Responsible AI needs guardrails: testing, security, red teaming, prompt injection defense
- •Top tools: GitHub Copilot, ChatGPT, and AI transcription/summarization (e.g., Granola)
