
ArchitectGBT
Your AI Architecture Expert
Hey Indie Hackers 👋
Integrated Gemini 3 Pro into ArchitectGBT's core recommendation engine, and the difference is night-and-day. The AI went from "pretty good" to "impressing Google's latest model good."
What I built:
A completely revamped dashboard that turns architecture decisions into a guided workflow. Instead of thinking in abstract patterns, users now see real trade-offs visualized, alternatives suggested, and risks highlighted before they ship.
The brutally honest part:
I over-engineered the UI my first time around. Redesigned it 3x based on beta feedback. The final version? So clean it feels like it should have been obvious from the start.
What I learned:
Gemini 3 Pro's reasoning depth is genuinely different from GPT-4
Users want explanations as much as recommendations
Most architects skip tools because they're too slow—speed matters more than perfection
What I need: 50-100 beta testers. Developers, architects, tech leads—anyone making architecture decisions at scale. You get free lifetime access in exchange for brutal feedback and bug reports.
Hey everyone,
I wrote a blog about how I built ArchitectGBT while struggling to pick an LLM for AllPub. The story goes:
Building AllPub → Need to pick an LLM → 3 hours of browser tabs and spreadsheets → Build ArchitectGBT → Write blog about it → Use AllPub to publish everywhere.
Here's the meta part: I wrote the blog in Notion, published to Dev.to, then used AllPub to cross-publish the exact same post to Hashnode instantly.
Same content. Two platforms. One click.
That's exactly the problem AllPub solves for creators.
If you want the full story about building ArchitectGBT:
1) https://dev.to/pravin_niceguy/architectgbt-how-to-avoid-costly-mistakes-when-selecting-an-llm-33p0
2) ArchitectGBT: The Secret to Smart LLM Decisions Made Easy
If you want to stop copying-pasting blog posts across platforms: https://allpub.co
Both problems are real. Both solutions are live. Both are in beta and need your feedback.
Curious what you all think about this "build a tool to solve a problem, then use another tool to share about it" meta situation 😅
#BuildInPublic #AllPub #ArchitectGBT
1 Like
Comment
Hey
I built ArchitectGBT because I was tired of wasting time comparing 40+ LLMs every time I started a new project.
What it does:
- You describe your use case in natural language
- We analyze it and recommend the best models ranked by speed, cost, and accuracy
- You get instant pricing breakdown + match confidence score
- Feedback system so we iterate based on YOUR needs
Current state:
- Core recommendation engine ✅
- Real-time pricing data ✅
- Feedback feature ✅
- Example prompts (code completion, chatbots, document analysis) ✅
Currently 3 promts are allowed for free tier.
What I need:
Your honest feedback. What's missing?
What would make you use this regularly? Is the ranking accurate for your use cases?
Try it here: https://architectgbt.com
Early access + beta testers get added benefits. Would love to hear what you think!
#BuildInPublic #AI #LLMs
1 Like
Comment
I'm building ArchitectGBT and want to validate if this is actually a problem people care about.
The Problem
Every time I start an AI project, I waste 5+ hours choosing the right model:
- Should I use Claude, GPT-4o, o3-mini?
- Which one is the cheapest?
- Which one is fastest?
- What about context window, latency, and reliability?
Then I pick one. Deploy. And realize 2 weeks later I picked wrong.
Last time: Last month i lost a good amount using GPT-4o when o3-mini would've worked perfectly.
I know I'm not alone.
My Solution (Idea)
ArchitectGBT:
- You describe your project
- AI tells you which model is best for YOUR use case
- See exact costs + performance tradeoffs
- Get starter code to deploy immediately
60 seconds instead of 5+ hours.
The Ask
I'm validating this idea with a simple landing page.
If you think this solves a real problem for you, join the waitlist.
Your email = signal to me that this is worth building.
Questions for You
1. Is this a problem you face?
2. Would you actually use this?
3. What would make you sign up?
Drop your honest thoughts in the comments 👇
2 Likes
3 Comments
3 Comments
-
1
Picking the wrong model wastes more time and money than people admit. The part I’m curious about is how deep your recommendations go. Most projects change once you hit real data or edge cases. Can it handle that or is it more of a starting point tool?
-
1
Hi Bill i have implemented enhancements on the recommendations have you tired the tool now? thanks
-
1
Thanks for the great feedback! You're absolutely right - projects evolve and edge cases matter. Right now, ArchitectGBT is designed as a starting point tool to help you narrow down options quickly. However, I am actively working on deeper features like:
Iterative recommendations (refine based on your feedback)
Real-world usage tracking (update recommendations as your project scales)
Edge case analysis (prompt for specific scenarios you might encounter)
Your insight about model selection being costly is exactly why I am building this. Would love to hear more about what specific features would make this most valuable for your workflow.
-
About
ArchitectGBT exists to remove that technical and mental barrier, so you can go from “Which model do I use?” to actually building your product in minutes, not weeks. I want indie hackers to ship faster.


Comment