Built This Week/AI Tools

AI Tools

Bolt

Bolt was our fastest way to get a front end on screen in 2025, and we still think of it for quick demos, but every serious build we started there moved to Cursor or Claude Code.

Our verdict on Bolt

Bolt (bolt.new) was where we started a lot of our early builds in 2025, because nothing got a working front end on screen faster. It is a good first step, not a finishing tool: every larger app we started in Bolt ended up in Cursor. By April 2026 we had mostly stopped using it, and quick demos moved to Google AI Studio.

What we built with Bolt

  • Mate (July 2025): Jordan's internal directory and culture app for our remote team, with profiles, tenure, off-site badges, interests, clubs and search. He scaffolded it in Bolt, including the left sidebar, then finished it in Cursor on Supabase. The whole thing took a few days over a weekend, and the team uses it all the time.
  • Daily Check-in (August 2025): Sam's digital stand-up. Each team submits priorities and blockers, and every morning an AI summary, blockers first, lands in a founders' Slack channel. He built it in Bolt between meetings, about two hours in total, and the day before recording it caught that the sales team had run out of credits for a tool.
  • Solopreneur finance dashboard for Collective (September 2025): Jordan built the first version in Bolt "really, really quickly", then moved it into Cursor and used OpenAI Codex to wire up the APIs, with Gemini doing the forecasts.
  • Make It Rain mini-game (September 2025): a three.js browser game where money falls and you collect it, with a timer and best score. Built entirely in Bolt. Jordan's top score on the show was 8,950.
Yeah. So I think I built the first version of this using bolt.new, which is an application that allows you to build front ends.

— Jordan Metzner, Episode 5

How we use Bolt

  1. Use it for the front end first. Jordan's rule from Episode 1 (at 20:04): Bolt or Lovable for product demos and front ends; Replit when you need an API, backend, database or auth.
  2. Replicate something you already have. Sam's first version of Daily Check-in was a replica of the stand-up form the teams already filled in, so nobody had to learn a new flow.
  3. Scaffold the layout, not the whole app. For Mate, Bolt laid out the shell and sidebar. That is the part it does fastest.
  4. Connect Supabase from inside Bolt. Bolt integrates with Supabase directly, so data and authentication live there from day one.
  5. Add the backend jobs around it. Sam used Supabase edge functions to summarize check-ins with AI, prompted to put blockers first, and a GitHub cron job to post the summary each morning.
  6. Move to an IDE when it gets serious. Take the code out of Bolt and finish it in Cursor, as Jordan did with Mate and the Collective dashboard.
  7. Ship, then iterate from feedback. Daily Check-in went live on a Monday, and Sam shipped a v2 from what the teams asked for.
And, you know, now I'm learning how to build edge functions and cron jobs, and it's all pretty easy with AI. Cumulatively, probably took me two hours total.

— Sam Nadler, Episode 7

Where Bolt falls short

  • Bigger builds. In August 2025 Jordan asked Sam if he would start his next prototype in Bolt now that he knew Claude Code. Sam said Bolt is "so fast", but for something as big as Sam's ATS he used Claude Code.
  • Adding AI to the app. In October 2025 Jordan explained that to put AI inside a Bolt or Lovable app you usually need a backend, an API choice and your own keys. Google AI Studio has Gemini built in.
  • Prototype to production. When Jordan showed the Collective dashboard, Hooman Radfar pointed out that getting from a fast prototype to production is the hard last mile.
  • Staying power. By April 2026 Sam said the tools we built on Bolt and Lovable were ones "we're not really using anymore".
  • What we did not test. We never discussed Bolt's pricing or plans on the show.
It depends on the use case, but, you know, Bolt's so fast. But I this was this is a pretty kinda big prototype.

— Sam Nadler, Episode 9

Bolt compared

  • Bolt vs Claude Code: Bolt for something small, quick or a landing page; Claude Code for bigger prototypes.
  • Bolt vs Replit: Bolt for front-end demos; Replit when you need a built-in backend, database or auth.
  • Bolt vs Google AI Studio: AI Studio is a step up for AI apps because Gemini is already wired in.
  • Bolt vs Figma: tools like Bolt, Lovable and v0 cut our Figma use, though Jordan still calls Figma the top design tool.
  • Bolt vs Lovable: we never tested them head to head. On the show they always sit in the same bucket, and Adir Ben-Yehuda of Autonomy AI described his app's prompt screen as something you know "from Bolt or Lovable".

Episodes featuring Bolt

  • Episode 1: when Jordan picks Bolt over Replit.
  • Episode 2: Bolt and similar tools cutting our Figma use.
  • Episode 5: Mate, scaffolded in Bolt and finished in Cursor.
  • Episode 7: Sam's Daily Check-in, built in Bolt with Supabase.
  • Episode 9: Bolt or Claude Code for the next prototype.
  • Episode 14: the Collective dashboard and Make It Rain game.
  • Episode 19: why AI Studio beats Bolt for AI-powered apps.
  • Episode 38: how we pick a building tool now.

FAQ

How do you use Bolt?

We used bolt.new to scaffold the first version of an app's front end, connected it to Supabase, and then either kept building there for small tools or took the code into Cursor to finish it. Sam built Daily Check-in this way in about two hours total.

Can a non-developer build a full-stack app with Bolt?

Sam, who says he is by no means a developer, built Daily Check-in in Bolt with Supabase edge functions and a GitHub cron job that posts an AI summary to Slack every morning.

Bolt vs Lovable: which is better?

We never ran them head to head. On the show we always put them in the same bucket: front-end product demos and quick prototypes, not apps that need a real backend.

Is Bolt good for big projects?

Not in our experience. Sam would still try Bolt first for something small, quick or a landing page, but used Claude Code for a bigger prototype, and Jordan moved his Bolt builds into Cursor to finish them.

Do you still use Bolt?

Not much. In April 2026 Sam said the things we built on Bolt and Lovable are tools we are not really using anymore, and for quick demos we now usually open Google AI Studio.