Will Go Thrive or Fade in the AI Era? A Deep Dive of GopherCon 2025 Roundtable

Go's New Opportunities: The Bedrock of AI Infrastructure

When asked what Go can offer that Python and other languages cannot, the panel unanimously highlighted its production‑grade reliability and concurrency capabilities. Samir Ajmani noted that Go’s rise was driven by the cloud‑native boom and that AI is creating a “second cloud‑native opportunity.”

Current state: AI/ML infrastructure heavily relies on Python for rapid prototyping.

Pain points: Scaling prototypes to production introduces high‑concurrency inference, complex agent orchestration, and protocols like MCP, where Python’s dynamic nature and performance limits become apparent.

Go's position: Go’s native high‑concurrency model, static type safety, and distributed‑system pedigree make it an ideal choice for AI production infrastructure (serving, orchestration, agent protocols).

Katie Hawkman shared a real‑world example: during a hackathon she chose Go over TypeScript to implement an MCP Server because Go’s code was easier to read and maintain for complex protocol logic.

David Soria Parra (Anthropic) observed that Go is currently one of the languages most frequently generated by AI, underscoring its advantage.

While Python may remain the “training language” for AI, Go is poised to become the dominant “runtime language.”

Career Anxiety: Will AI Replace Us?

The panel agreed that AI is merely a new productivity tool that changes work patterns but ultimately enhances human value.

Samir Ajmani: Future software will be assembled from components, requiring engineers who understand system design, security, and reliability. The barrier for simple code generation rises, but opportunities expand for those with deep engineering skills.

Jaana Dogan (Google): Faster coding with AI lets engineers spend more time “connecting the dots,” moving from “screw‑driver” roles to system‑architect roles.

David Suryapara (Anthropic): Pure coding skills (e.g., memorizing APIs) may devalue, while core engineering abilities—decomposing complex requirements, designing distributed systems, handling edge cases—become even more indispensable.

Katie Hawkman: The most interesting part of coding is not the act itself but progressive delivery, UX design, and performance optimization—areas AI cannot fully replace.

Ian Cottrell: Historical productivity tool leaps (assembly → C, IDEs → auto‑completion) have always increased demand; they raise expectations rather than eliminate developers.

Advice: don’t try to master every AI tool; pick one (e.g., Cursor or Claude Code) and integrate it deeply into your workflow.

Rational View: Compute, Energy, and Responsible AI

The host asked how to rationally assess AI’s massive compute and energy consumption, especially after blockchain’s criticism for high energy use.

Samir Ajmani (Google): An experiment integrating MCP support into the Go language server (LSP) showed higher task completion rates, lower latency, and nearly 50% reduction in token consumption, demonstrating that solid engineering tools can cut AI runtime costs and carbon footprint.

Jaana Dogan (Google): We are at an early stage of optimization; like past database tuning, inference efficiency—through caching, quantization, and specialized hardware—will be the next focus.

David Suryapara (Anthropic): Small, distilled models fine‑tuned for specific domains (e.g., code generation) can balance performance and cost, avoiding reliance on large, expensive models.

Responsible AI is not just an ethical imperative but an engineering necessity: using fewer tokens to do more aligns with Go developers’ expertise in resource optimization.

Practical Survival Guide: Filter Noise, Return to Fundamentals

Start Small: Don’t be intimidated by AGI hype; begin with simple agents or an MCP server in Go, even as a beginner.

Focus on Determinism: AI models are probabilistic, but Go’s static analysis, testing toolchain, and type system provide the determinism needed for reliable systems.

Solve Real Problems: Apply AI only where it truly improves efficiency (e.g., automated documentation, complex log analysis) rather than forcing AI into every feature.

Conclusion: Go Community’s Greenfield Moment

The strongest signal from the roundtable is optimism. We are at a moment comparable to the pre‑Docker era of 2013, with AI‑centric “Kubernetes” and “Prometheus” still unwritten—an open greenfield for Go developers.

“If I want AI to truly transact with the real world—like ordering a pizza and having it delivered—we need massive, reliable infrastructure. Go is the perfect language to build that layer.”

Go engineers are encouraged to leverage their concurrency models and engineering wisdom to lay the steel foundation for the AI era.