Adam Larson (GosuCoder)

Activities

historical

  • Software engineer with a focus on tooling and developer workflows.
  • Founder / CTO of Railon (listed in conference/speaker materials) and has held senior engineering roles; some listings indicate a transition to principal staff engineering at Mailchimp (see Sources).
  • Gained visibility through deep, practical testing of AI coding models and agents.

present

  • Runs the YouTube channel “GosuCoder”, producing frequent, hands-on reviews, bench tests, and demonstrations of AI coding assistants and coding-focused LLMs (Claude, Grok, Gemini, Qwen, etc.).
  • Produces technical content that emphasizes realistic evaluation: plan-driven prompting, agent orchestration, and codebase-level testing rather than synthetic benchmarks.
  • Regular participant/guest in industry podcasts and conference talks (e.g., DevCon Fall 2025, Rate Limited podcast appearances).
  • Active on X/Twitter as @GosuCoder and maintains a LinkedIn presence (see Sources).

Practical usage examples (how to learn from his work)

  • Follow his plan-mode and “plan-first” walkthroughs to adopt a structured, task-planning approach to using LLMs for engineering tasks (useful for building reliable agent workflows).
  • Watch his side-by-side comparisons (e.g., Claude Code vs. Gemini vs. Grok) to understand real-world tradeoffs between model cost, latency, and code correctness.
  • Use his codebase-indexing and semantic-search demos (RooCode, Augment Code, Cursor) as examples of integrating LLMs into an IDE or development pipeline.
  • Study his evaluations when designing your own model-evaluation harness: he stresses that harness and toolchain choices can change perceived model rankings.

Connections to other people and companies

  • Railon (founder / CTO listed in event materials).
  • Mailchimp (listed as principal staff engineer in conference listing).
  • Appears on or contributes to developer/AI industry podcasts (e.g., Rate Limited) and speaks at conferences (DevCon Fall 2025).
  • Engages with AI tooling companies and projects through product reviews and testing (Cursor, RooCode, Claude, Augment Code, various model providers).

Expectations for the future

  • Will likely continue to focus on rigorous, pragmatic evaluations of coding LLMs and agent workflows.
  • Expected to keep producing content that highlights where coding agents are genuinely useful vs. overhyped — useful for teams adopting AI-assisted development.
  • May publish more structured benchmark suites or talks that formalize his evaluation methodology (he has indicated work on more robust evaluation frameworks in talks).

Interests

  • Practical application of LLMs to software engineering tasks: code generation, refactoring, debugging, and agent orchestration.
  • Tooling and editor integrations that make LLMs productive for daily developer workflows (e.g., Cursor, Zed, RooCode integrations).
  • Improving evaluation methods for coding models and encouraging reproducible, realistic testing.

Sources