The State of Startups 2025 - Key Takeaways
Supabase recently surveyed over 2,000 startup founders and builders to uncover what's powering modern startups: tech stacks, GTM, and approach to AI.
Some results that struck me:
AI coding tools are indispensable for startups, and not just @Cursor (27%) and @VS Code (25%). Vibe coding tools like @Lovable, @bolt.new, and @v0 by Vercel are also common.
Most startups are already integrating models like @OpenAI (39%) or @Claude by Anthropic (21%), especially for semantic search, summarisation, and customer support. Half are building agents to automate real tasks, from onboarding flows to sales triage.
Tool discovery happens quite often via @YouTube or @GitHub. Physical event participation remains low.
Founders earn their earliest customers through networks (58%), communities, and inbound content. Paid acquisition rarely works early on, nor does performance marketing.
The hardest problems are still the oldest ones: customer acquisition, product-market fit, and complexity.
61% founders remain optimistic about the future.
Replies
minimalist phone: creating folders
I am surprised that there's no mention of TechCrunch in the product discovery. I usually go there + PH.
for context, participants are mostly technical (81%). 5% of them mention @Product Hunt as a place where they discover new products. I suppose TechCrunch is in the Other category (4%).
minimalist phone: creating folders
@fmerian TBH, I expected more for Product Hunt :D
likewise! we're biased tho
minimalist phone: creating folders
@fmerian Yeah, in our ideal bubble, everybody knows about PH :D
Super insightful roundup 👏
Loved seeing how AI coding tools and agents are shaping early-stage workflows, especially the shift toward automating real tasks like onboarding and sales triage. We’re seeing the same in QA automation: founders don’t just want efficiency, they want their time back to focus on growth.
Also agree on the takeaway around customer acquisition, communities and networks remain the strongest growth levers. Paid rarely cracks the early story.
Thanks for sharing this report, definitely bookmarking for our own roadmap learnings.
— Team OopsBot
Computer Using Agents by LLMHub
I am curious about the 4th point here, so currently we are building llmhub.dev which we just launched and are trying to find our small niche of customers who we can make content and show the value proposition, how do you suggest we make content or find customers for our tool here based on your experiences, would love to know your thoughts and also if you have any critiques about our product ;)@fmerian
4 and 5 make a lot of sense to be honest. Those will come last in being solved or assisted for by AI. You can get a lot of help in the building part but LLMs seem to be not as valuable when it comes to building actionable plans about finding customers.
WebCurate.co
The 5th one. Is there any AI tool to handle that fully yet? I wonder if any!
I'd be interested to see the correlation/causation between the growth of ai tools (native language driven or suggestive) and the drop in size in founder teams! Seems there are some big big apps being produced by individuals
Point 3, tool discovery, ...really? How is that happening? A bit more detail would help
for context, participants are mostly technical (81%). @YouTube and @GitHub are places where they hang out.
Documentation.AI
Any idea on how the tool discovery happens via GitHub. Is this only for open-source tools or for all tools? We are building Documentation.AI, It would be very relevant for us.
GitHub is where developers hang out the most. Definitely an acquisition channel for developer tools.
Random ideas:
Sponsor popular projects. Take @Mintlify and @Inkeep sponsoring @Fumadocs for example.
Contribute to them
Make sure your repos show up in GitHub SERP
Build free, open-source tools. Take @Wasp and @Raycast with resp. @Open SaaS and @Ray.so.
Hope it inspires!
I wonder if tool discovery shifts from YouTube and GitHub to ChatGPT/Grok/etc in the upcoming year or two.