--- title: "AI Week: Is Spec-Driven Development the Future of AI Coding?" description: "Tessl CEO Guy Podjarny explains spec-driven development: defining software through specifications instead of code. Learn the three stages of adoption and how Tessl's framework makes AI agents reliable for production use." canonicalUrl: "https://zuplo.com/blog/2025/10/01/spec-driven-ai-development" pageType: "blog" date: "2025-10-01" authors: "josh" tags: "AI" image: "https://zuplo.com/og?text=AI%20Week%3A%20Is%20Spec-Driven%20Development%20the%20Future%20of%20AI%20Coding%3F" --- Guy Podjarny, founder of Snyk, is now building [Tessl](https://tessl.io) to solve a problem every developer using AI tools has encountered: agents are powerful but can be inconsistent, especially when repeating tasks over and over. They hallucinate APIs, break existing functionality, and forget decisions that their human developer counterparts already made. The solution? Stop defining software through code and start defining it through _specifications_. _Watch the full conversation between myself and Guy Podjarny (Tessl CEO), as part of AI Week and the launch of Zuplo's new [AI Gateway](https://zuplo.com/ai-gateway) below:_ ## Why create a company like Tessl? Tessl is built on the belief that we'll move from defining software through implementation to defining it through specifications: capturing what you want to build along with constraints and requirements. This shift unlocks new possibilities: software that's easier to create, autonomously maintained, and adaptable to different contexts like changing business rules, programming languages, or runtime environments. ## Three Stages of Spec-Driven Development The transformation to spec-driven development isn't binary. It's a journey that evolves through three distinct stages, each building on the last: 1. **Spec-Assisted Development** provides agents with structured knowledge: coding standards, architectural decisions, API documentation. Unlike humans, agents will actually read the documentation you give them. This alone makes them significantly more useful. 2. **Spec-Driven Development** captures critical definitions as specifications that become the source of truth. Before making code changes, you first modify the spec, then apply the change. The specifications might describe your checkout process, core algorithms, or API contracts. 3. **Spec-Centric Development** is the endgame: comprehensive specs and tests make the code disposable. You can regenerate it whenever needed, adapt it to different contexts, or optimize for new constraints without worrying about accumulating technical debt. ## Incomplete Specs Are Actually _Better_ Unlike source code compilation, specs are intentionally incomplete. You define what matters (business logic, constraints, requirements) and let AI handle implementation details. This incompleteness makes specs more usable and allows generated code to adapt to different contexts. The key is specifying _what_ you want to achieve, not _how_ to build it. ## Tessl: Framework and Registry The recently released **Tessl Spec Registry** works like NPM for knowledge: a dependency system for versioned specification packages. It comes pre-populated with over 10,000 usage specs for popular open source libraries, solving the acute problem of agents struggling with libraries that are too new, too old, or too niche. You can also publish your own spec packs for internal practices or public use. Tessl Spec Resgistry is available now at: [tessl.io/registry](https://tessl.io/registry/) As well as this, they also hace **The Tessl Framework** which plugs into existing agents via MCP (Model Context Protocol), providing tools for creating specs, maintaining spec-first workflows, and generating aligned tests. It works with Claude Code, Cursor, Codex, Gemini CLI, and other MCP-compatible agents. Tessl Framework is currently in closed beta. You can join the waitlist at [tessl.io](https://tessl.io/). ## Guy's Prediction: Most Developers Won't Look at Code by 2027 Using [self-driving cars as an analogy](https://ainativedev.io/news/the-5-levels-of-ai-agent-autonomy-learning-from-self-driving-cars), Guy's prediction is that by end of 2027, developers working with agents won't look at code most of the time. Within 2026, most development will be at least spec-assisted. It isn't just rapid improvements in LLMs that's driving this. It's patterns and tooling that give agents the information they need to succeed and provide the guardrails to keep them trustworthy. For more insights like this, we recommend taking a look at the AI Native Dev community at [ainativedev.io](https://ainativedev.io). ## More from AI Week This article is part of Zuplo's AI Week. A week dedicated to AI, LLMs and, of course, APIs centered around the release of our [AI Gateway](https://zuplo.com/ai-gateway). You can find the other articles and videos from this week below: - Day 1: [AI Gateway Overview](/blog/zuplo-ai-gateway) with Zuplo CEO, Josh Twist - Day 2: [Is Spec-Driven AI Development the Future?](/blog/spec-driven-ai-development) with Guy Podjarny, CEO & Founder of Tessl - Day 2: [Using AI Gateway with LangChain & OpenAI](/blog/ai-gateway-with-langchain) with John McBride, Staff Software Engineer at Zuplo - Day 3: [Your AI Models Aren't Learning From Production Data](/blog/comet-ml-opik) with Gideon Mendels, CEO & Co-Founder of Comet ML - Day 3: [Using Claude Code with Zuplo's AI Gateway](/blog/ai-gateway-with-claude-code) with Martyn Davies, Developer Advocate at Zuplo - Day 4: [What Autonomous Agents Actually Need from Your APIs](/blog/what-autonomous-agents-actually-need-from-your-apis) with Emmanuel Paraskakis, CEO of Level250 - Day 4: [Using AI Gateway with goose AI agent](/blog/ai-gateway-with-goose) with Martyn Davies, Developer Advocate at Zuplo