Built This Week/Explore

Explore

AI Tools We Use

Every AI tool we have actually used on the show, with what we built, how we use it and where it falls short.

By Jordan Metzner & Sam Nadler · Updated October 2026

45 pages, each built from what was actually said and built on Built This Week. Want the full episodes? Browse every episode.

AWS AmplifyAWS Amplify is the boring last step in our stack: once Sam has an app working in Claude Code or Codex with Supabase, Amplify and GitHub get it live on…BoltBolt 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 move…ChatGPTChatGPT is our go-to for writing PRDs and prompts and the API brain inside several internal tools, but for code and visual output we kept picking Clau…ClaudeClaude is our pick for turning files into dashboards and visual web tools in chat, and its Opus and Fable models set the bar for coding quality, but c…Claude CodeClaude Code is the fastest way we know to go from a PRD to a working app, especially with several sessions running in parallel, but we hand finished w…Claude CoworkClaude Cowork is Claude Code with a friendlier face: point it at a folder of docs and it plans, proposes trade-offs and shows progress; once you want …Claude DesignClaude Design feels like a UX designer you can chat with; Sam, who calls his Figma skills awful, used it to redesign an internal app and ship it throu…CursorCursor was our main IDE through most of 2025 and the place we finished apps started in Bolt, but by 2026 our most advanced work had moved to Claude Co…ElevenLabsElevenLabs is the voice tool we and our guests reach for: Jordan used it to narrate a 65-page slide deck in Remotion, and Breeze's founder called it o…Envato AIWe tried Envato AI live for Instagram ad images and none were usable, because every image mangled its text; we still use Envato for stock content, not…FigmaFigma is still the top web-based design tool in our view, but since AI tools can design an interface from a prompt, we open it less and less, and Clau…FLUXFLUX was a dependable image model inside our 2025 internal tools and its Kontext Pro edit model kept a face recognizable across chained edits, but it …GammaGamma is the fastest way we know to get a deck or one-pager 65 to 80 percent done; you still review the text and fix images before it goes out.GeminiGemini is our pick when an app needs AI built in or has to read a lot of data, and guests praise its speed and cost, but for coding we have kept reach…Google AI StudioGoogle AI Studio is where we build quick demos and throwaway apps, because Gemini is already built in, but for real software we still move to Claude C…Google AntigravityGoogle Antigravity is the editor we open to test new Gemini models for free, but in our use it was less polished than Cursor and never became our dail…Google VeoVeo 3 was the step up in AI video that made us build our own studio, and Veo 3.1 Fast beat Sora 2 in our side-by-side test, but expect slow, pricey re…gpt-ossgpt-oss-120b felt close to GPT-4.1 in our one live test and ran in seconds on GroqCloud; its real promise is near-zero cost for always-on agents.GranolaGranola is the AI note taker we actually kept: it doesn't join the call, it lets Sam stay present through 15 interviews a week, and its transcripts no…GrokWe have tested Grok as one model among several, and guests respect how fast xAI caught up, but as of 2026 we have not found a high-quality use case to…GroqCloudGroqCloud is where Jordan runs OpenAI's open-source models when he wants speed; in our one live test it turned out ten drafts in about five seconds.JiraWe didn't review Jira itself; we fixed the worst part of using it. A TPM on our team built RADAR, which turns meeting transcripts into suggested Jira …LeverLever was the applicant tracking system behind our AI hiring tools; we kept it as the system of record and sent AI scores and interviews back into it …LovableWe treat Lovable as a fast way to get an idea on screen, not a place to ship production software, and by 2026 we had mostly stopped using it.MapboxMapbox was the map layer for Tix LAX, Jordan's Los Angeles parking ticket map; the AI coding tools wrote nearly all of it, and the hard part turned ou…Meta Muse ImageSam found Meta's Muse Image feels pretty similar to Nano Banana or OpenAI's image model; what stood out was that it can remix anyone's public Instagra…n8nn8n is the workflow tool we picked for AI automations; the fastest way past its learning curve is to have Claude write the workflow JSON and import it…Nano BananaNano Banana is the image model we reach for when the subject has to stay the same across edits: product shots, costumed selfies and ad images, usually…NotebookLMNotebookLM is the tool we reach for to turn plain documents or a research question into learning material: infographics, slide decks, audio overviews …OpenAI CodexOpenAI Codex on extra high reasoning is slower and less fun than Claude Code, but it writes cleaner code and finds the bugs Claude Code misses, so it …PerplexityWe use Perplexity as the research step inside automations rather than as a daily chat app; it feeds company research into an n8n sales workflow, and i…Polygon.ioPolygon.io was the small but necessary piece of Jordan's AI trading bot: a free price API he polled every 15 seconds so he didn't burn through his bro…PomelliPomelli turns your website into a 'business DNA' and then into on-brand social campaigns in minutes; it is great for quick, elegant ideas, but some im…PublerPubler is how our social team gets 8 to 19 posts a day out across several brands without spending all day uploading; we switched from Buffer and have …RemotionRemotion is a React video library that AI coding tools can drive, so people with zero editing skills can build branded, repeatable videos; the catch i…ReplicateReplicate is how we plug new AI image and video models into our own apps: copy the API sample and schema, paste them into Replit, and the model usuall…ReplitReplit is Jordan's pick for small tools that need a backend, database and API keys without setup; Sam likes it least, and for big software we now use …SeedanceSeedance was the fast option in our internal AI video studio: ByteDance's model made a clip in moments, while Veo 3 could take up to ten minutes and c…SlackFor us Slack is where AI tools report back: Sam's daily check-in summary posts blockers to a founders' channel every morning, and Jordan mocked up a c…SoraSora 2's Cameo feature, putting your own face in a video, was the most fun we had with AI video in 2025, but in our side-by-side tests Google Veo stil…StrudelStrudel is a JavaScript live-coding library for music whose code is easy for people to read and for LLMs to write, which is why Jordan could build an …SunoSuno is the fastest way we know to turn a text prompt into a finished, good-sounding song; we made our show's jingle with it, though Jordan doubts AI …SupabaseSupabase is the default backend under almost every internal tool we vibe code: auth, the database and edge functions, simple enough that Sam, a non-de…WebflowWe moved our company websites from WordPress to Webflow to spend less time on maintenance, and found the SEO just as good, if not better; its CMS API …Wispr FlowWispr Flow is the dictation app we both use daily: it is far more accurate than Apple's built-in dictation and lets us give AI tools long, detailed pr…