Show HN: I built an AI senior architect – vibe coding meets system design(sysdesai.com)
sysdesai.com
Show HN: I built an AI senior architect – vibe coding meets system design
https://www.sysdesai.com
2 comments
The framing of "architecture part of vibe coding that usually gets skipped" is the key insight here. Most vibe coding tools optimize for feature generation — they'll happily build you a rate limiter, an authentication flow, and a database schema, each in isolation, without any reasoning about how they fit together or whether the architectural decisions are coherent.
The step-by-step approach with human push-back opportunities is meaningful because it keeps architecture as an explicit conversation rather than an implicit output. When you can say "I disagree with this tech choice" at the architecture phase rather than discovering the mismatch after the code is generated, you maintain the accountability that vibe coding workflows often lose.
This connects directly to the Agile Vibe Coding Manifesto's principle that "architecture guides and constrains generation." The manifesto argues that architecture shouldn't emerge from accumulated vibe coding sessions — it should be the input that shapes what gets generated. Tools like this that make architectural reasoning explicit are building in the right direction.
Interesting product for teams trying to maintain system design rigor in AI-assisted workflows: https://agilevibecoding.org
The step-by-step approach with human push-back opportunities is meaningful because it keeps architecture as an explicit conversation rather than an implicit output. When you can say "I disagree with this tech choice" at the architecture phase rather than discovering the mismatch after the code is generated, you maintain the accountability that vibe coding workflows often lose.
This connects directly to the Agile Vibe Coding Manifesto's principle that "architecture guides and constrains generation." The manifesto argues that architecture shouldn't emerge from accumulated vibe coding sessions — it should be the input that shapes what gets generated. Tools like this that make architectural reasoning explicit are building in the right direction.
Interesting product for teams trying to maintain system design rigor in AI-assisted workflows: https://agilevibecoding.org
Happy to answer any questions about how it works
So I built SysDesAi. It started as a tool for myself. Part AI architect, part learning platform, part engineering news source. You describe what you want to design and an AI walks through it with you step by step: requirements, capacity estimates, API design, data models, architecture diagrams, sequence flows, scaling bottlenecks. Basically the architecture part of vibe coding that usually gets skipped.
It's not a one-shot generator. You can push back at every step, ask questions, or take the conversation in a different direction. If you disagree with a tech choice the AI made, you say so in chat, and there's a button to regenerate just that step with the new direction. Or if you don't care and just want the full design, there's an autopilot mode for that.
It ended up growing beyond the design tool:
- 10 phases per design (requirements, scope, estimates, API, data model, architecture diagram via Mermaid, sequence diagrams with a custom SVG renderer, scaling, summary) - Two AI modes: Fast (Gemini 2.5 Flash) and Gemini 2.5 Flash Thinking - Google Search grounding on the architecture and scaling phases, so the AI recommends tech based on what's current, not just what was in its training data - Gallery with 48+ designs you can browse, fork, comment on, or ask the AI about - Free course: 11 modules, 60+ lessons, quizzes, certificates - News feed pulling from 35+ engineering blogs (Netflix, Uber, Cloudflare, AWS, etc.). You can read an article and generate a design based on it
Some technical decisions I found interesting:
- Designs aren't generated in one shot. You clarify requirements first, then the AI builds a step plan, then generates each step with full context of everything before it. The difference vs. a single big prompt was pretty obvious once I compared the two. - The architecture and scaling phases use Google Search grounding. Without it the AI would recommend outdated patterns or miss newer alternatives. With it, the recommendations actually track what people are using now. - When you ask the AI to change something in chat, it detects that and shows an "apply changes" button. One click regenerates that step. Downstream steps get marked as outdated so you know what needs a refresh. - I built a custom SVG sequence diagram renderer because Mermaid's sequence diagrams can't do interactive step reveal, clickable step bands, or zoom/pan. Spent more time on this than I'd like to admit.
Pricing: 50 free credits on signup. You can bring your own Gemini API key and use it for free, unlimited. Or buy credit packs ($5/50, $20/250) if you don't want to deal with API keys. Course, gallery, and news feed are free, no credits needed.
https://sysdesai.com
Not sure how useful this is to anyone besides me, but I'd like to hear what you think. What works, what doesn't, what would you do differently.