What have you built with LLMs?

I built a multiplayer chatroom where all messages are transformed by an LLM (e.g. into pirate speak or corporate jargon):

https://impersona.chat/

I also built this incremental clicker game where you split words ad infinitum (like Infinite Craft but in reverse):

https://lantto.github.io/hypersplit/

I built Audiala [1][2][3], which is a platform that creates virtual tour guides that goes with you anywhere in the world. While the content is mostly static now, the pipeline is there to make it entirely personal for each user, once I figured out how to deal with the API costs.

The next steps are to adapt the content based on realtime user feedback or queries like "What happened on this street during the French Revolution?" or "Tell me more about this building’s history." as well as offering step by step itinerary suggestions.

I believe it's now possible to build an Augmented Reality solution, where you leave your phone in your pocket, put some airpods and you listen or converse with your personal tour guide and fully immerse yourself in history.

[1] https://audiala.com

[2] https://apps.apple.com/us/app/id6446038181?uo=4

[3] https://play.google.com/store/apps/details?id=com.audiala.au...

Great question!

I suck at web design, so I built a novel tool [0] that streamlines this flow:

webpage —> screen capture —> LLM prompt —> design feedback

It’s a browser extension that lets you capture a portion of the webpage —> the image gets sent to an LLM with a prompt —> the LLM gives design feedback inside the extension

As far as I know, this is the only tool I’ve seen that does this, and it’s completely free. It’s called Design CoPilot, and I haven’t done any real marketing yet as it’s in Beta.

Before I built this, I would manually screen capture components, then go to ChatGPT and drop the image, write a prompt telling ChatGPT it’s a design expert and I want design feedback, and then read its reply and implement the feedback.

I’d greatly appreciate if you try the tool and leave a review!

[0] - https://chromewebstore.google.com/detail/design-copilot/hgal...

I run a design studio. We make story maps a lot. It's a pain.

So I built a tool called Uiino. It lets you make story maps with plain text. Later, I added AI. Now it can generate maps instantly.

It's helped tons of people map out their apps. Gives them a clear bird eye view of their apps/flows.

Try → https://www.uiino.com (No signup required)

P.S. We just hit 10,000 website visitors. And 300+ monthly active users. It's all free. I'm pretty stoked about it.

I've built an LLM powered Anki clone - https://vocabuo.com. In my language learning journey, I've found that adding each word manually, with audio, image and a sentence example was just too tedious, so in Vocabuo, I add a word and LLMs + dalle + google tts generate the rest.

Currently it supports English, German and Spanish, but I plan to add more languages in the future.

As a bonus you can also add words from yt videos and websites.

I've built an open source data analysis and visualization tool https://dataline.app

Now my product manager and customer support teams use it for most small questions / charts needed.

I built a native all-in-one AI client called BoltAI[0]

It’s one of the first AI clients I think. I’ve been working on it for 16 months and shipped more than 100 releases.

It started with the idea of invoking an LLM within the Apple Notes app. I hate going back-and-forth between the Notes app and ChatGPT web app.

So I built the “AI inline” feature where I can trigger a keyboard shortcut and it would send the prompt to OpenAI and get the answer back.

But this is quite advanced, not many users can use it.

So later I added the full Chat UI.

I don’t want to charge a subscription and manage AI tokens so I sell perpetual licenses with 1 year of update. Users bring their own API keys.

As an all-in-one client, it unlocks some interesting features that I didn’t expect in the beginning: the ability to switch between multiple AI services and models.

This allows me to use Web Browsing with Claude, for example.

I later added support for Function Calling and this has unlocked many capabilities: edit video with ffmpeg plugin (give instructions to the LLM and it would execute the ffmpeg command locally), search the web using advanced search engine such as Kagi, WolframAlpha, Analyze documents and source code similarly to Claude Project…

And finally I added the AI Command feature where you can prompt directly with the highlighted text. Imagine you’re reading an article and wanted to list key takeaways, you can press a keyboard shortcut, choose the command and it would show you a quick answer. You can build any command you want: from summarization to translation or grammar fix…

I also built another free app called ShotSolve[1]. It allows you to take a screenshot with a keyboard shortcut and quickly analyze it with an LLM.

During building BoltAI, a customer asked for the ability to analyze PDF documents so I decided to build another app for that use case called PDF Pals[2]

These apps allow me to quit my job and pursue solopreneurship full time.

[0]: https://boltai.com

[1]: https://shotsolve.com

[2]: https://pdfpals.com

I build a personalized newsletter / podcast that picks relevant news based on your preferences and comes every morning.

https://www.tailoredpod.ai

I've been coding more in C because ChatGPT makes it less tedious. Any simple command line tool I will usually write in C now instead of Python because the end result is way faster.

I'm building a simulated fantasy town backed by an LLM. Visitors to the website can vote on what the hero does next, and everyone in the town will react naturally to all of the heroes actions.

Unlike previous efforts in this space, the technique I am using consumes very little context, and I'm hoping to get it running on consumer GPUs.

I made https://whiteboard.lasky.io, which is basically if you gave an LLM a marker and told it to solve math problems.

Also made https://resgen.app to help candidates tailor resumes

I was fed up with long, search-optimized introductions for online recipes and decided to make a page with recipes that have infinite introductions:

https://a.mancato.nl/recipe

It uses WebLLM so I don't have to pay/send data to an AI provider.

I built a service that turns entire websites into structured output: https://sitewideai.com

You enter a starting URL, describe the data you want in a prompt, the AI suggests columns for the output spreadsheet which you can customize, and then goes off and turns the website into structured data into a CSV file.

It also supports limits, you can say for example "visit at most 100 pages" and it will stop after 100 pages.

It was easier said than done to get prompts working as intended and the crawler to focus on the most relevant URLs first. As always, the final 20% end up taking up 80% of the development time.

I built an error correction model for people with significant disabilities who can’t speak and poor finger control to write out their message. https://willwa.de/2024/01/12/rapid-development-in.html app: look for dragtospeak on AppStore and model https://huggingface.co/willwade/t5-small-spoken-typo

Not built but planning to build a government knowledge repository with prompt query for intuitive and interactive understanding of government agencies operational documents including the Acts, Policies, SoP, Guidelines, Memo, etc based on LLM and RAG that can run locally using open source LLM model.

It's similar to this AI-driven chat assistant for ECE 120 course at UIUC:

https://www.uiuc.chat/ece120/chat

I built Heuristica for learning and research using concept maps powered by LLMs. It is a different kind of interface for using LLMs. Instead of having linear conversations, you can have visual explorations that let you structure, organize, and retain knowledge much better.

I have been working on it on the side for the last 17 months.

https://www.heuristi.ca/

I've built SourceSailor A Node Based CLI [1], an open-source CLI [2] that leverages LLMs (OpenAI, Anthropic, and Google Gemini) to analyze and document codebases. Key features:

- Quick project structure and dependency analysis - Smart file exclusion (respects .gitignore) - Tailored analysis based on user expertise - README generation from codebase insights

Recent updates include enhanced CLI aesthetics and multi-LLM provider support. Docs and additional tests are in progress.

Curious to hear your thoughts or feature suggestions! (And if you visit, I'd love to know if you came from this HN thread )

[1]:https://www.npmjs.com/package/sourcesailor?activeTab=readme

[2]:https://github.com/PrashamTrivedi/SourceSailor-CLI

Not much yet except a little email wrapper around the top LLMs. Now I can send an email to claude@goodbot.club or gemini@goodbot.club and get an answer in a few seconds.

I found in a recent project that it was interesting to compare the answers between GPT, Gemini, Claude, and Mistral. I have better results by combining the parts I like in each answer. It's also easier to ask follow up questions directly from my email client regardless of which device I'm currently using.

I built the app in a weekend with Ruby on Rails and AWS SES. I'm now wondering if it could be useful to other people? It's already open source and I could probably make the hosted version free by allowing users to provide their own API keys.

https://github.com/vinc/goodbot.club

We built hoop.app which uses LLMs to detect your tasks and work priorities from your chat messages, meeting transcripts and emails.

The goal is to take all of the busy work out of task management so you can always know what your focus should be on and not waste brain power worrying about missing something important.

I made a financial chat bot. It can call some APIs, perform portfolio weight optimizations, run optimized weights out of sample, stores your past works, presents correlation heatmaps and shows equity curves. It can also hallucinate endlessly about taking over the world.

I'm building a website (https://pocket-quant.com) where you can use LLMs to ask questions about past earnings calls.

You can see the earnings report and calls side by side and ask questions about them.

I'm also working on adding a live calls feature that can trigger workflows. So you'd pose a question before the call, define some answer structure (using structured outputs) and then connect it to some kinda of integration (e.g. invoke a lambda, slack/discord message, etc.)

As the call happens in real time I transcribe it and run the output by the llm to answer your question.

I see it being useful for fed talk, earnings calls, and open to ideas for other data sources!

- tool to analyze and explore unity3d build size reports

- tool to manage examples dataset for another ai application

- macos transcribe app that allows me to hotkey to speak to notion page, with a little status bar menu to select pages

All relatively trivial but I'd never bother to create them manually

I built a website to help people find summer camps for their kids.

https://campwing.camp/

Right now it’s just what i got up in a hurry last year, covers Seattle area only. Next year will cover SF and LA at least (and have a lot more features).

The data is quite granular (each row is one batch, say x age group at y location with z theme) and I do a lot of structuring data with prompts and now OpenAI’s Structured Outputs. I think the data gathering would have made this cost prohibitive otherwise.

I’m not engineer by training or work experience, so this also would have been impossible for me to build without ChatGPT’s help coding.

I'm a medical student with FOMO and who loves AI stuff.

I couldn't keep up with my news so I made the perfect summarizer that goes through the thought process of the author : https://github.com/thiswillbeyourgithub/WDoc

I needed an AI based system that go through my anki cards, but might as well make it able to read dozens of file formats. Now I can put entire medical youtube playlists, conferences, anki databases, hundreds of PDFs and ask a single question across all of them at once .

It's both the same project

I have a couple. A fun one is catchingkillers.com. Here you are a private investigator who is trying to find out who committed a murder. Two stories are active. The third, bathtub one, isn't done.

I made an encyclopedia where every page is generated by LLM on demand, so there's no such thing as a topic that doesn't have a page. It's like a real life hitchhiker's guide to the galaxy.

https://incoherency.co.uk/encmech/

https://incoherency.co.uk/blog/stories/encyclopedia-mechanic...

I built a new fog of war system for my virtual tabletop, it uses shaders, I know nothing about shaders: https://bonfiretabletop.com/

You'll have to create an account to see it, but I've posted a preview on Twitter: https://x.com/BonfireVTT/status/1833988532978823259

We've built Cradle, a framework that leverages such models to perform complex computer tasks via the same general interface humans use: screen as input and keyboard & mouse operations as output.

It works both on regular software and in complex games like RDR2. And it doesn't cheat by using any game-/software-specific API, nor accessibility calls, nor DOM trees. :)

https://baai-agents.github.io/Cradle/

And we're still evolving it.

Made a domain name generator: https://namebrewery.com/ Used the project to learn a bit of SvelteKit.

I made an extension [1] to summarise terms and conditions of your current webpage. After reading the recent news related to disney+, the idea struck me what all unusual clauses websites put in their T&Cs.

It currently doesn't work on some sites.

[1] https://chromewebstore.google.com/detail/tc-summarizer/kjdfi...

I use LLMs to generate meal plans for my app https://getweeklio.com. You can also generate recipes for any of the meals. Another functionality I want to add is being able to get a list of ingredients to make a shopping list. Right now al of it is using chatgpt4o. I put a cache in place initially but with the cost/speed of the new models it's becoming kind of unnecesary.

Gotta love how one response here is we built a candidate filtering tool and the other is we built an AI tool to solve interviews.

Sounds like job search in future is going to be wild(er)

I've been playing around with LLMs on the side, using ChatGPT to categorize comments on WhoisHiring by location, roles, tech stacks, and summarizing them. I'm still experimenting with phi3, llama3, and gpt-4 to see what works best for me. It's just a weekend project, but it's been a fun learning experience so far!

https://weworkjs.com/about

I built a powerful medical news survey report writer:

https://github.com/impredicative/newssurvey

It is to be extended to other sites as well. If you read the generated sample reports, you will see how it has potential to change lives.

Unlike much of the commercial stuff that people make, which almost nobody ends up using, the best part is that it is open source.

built https://chatcraft.org over the past 1.5 years. I felt chatgpt was too limiting from being proprietary. Can run js,ts,python inline. I find it to be a much better tool than copilot for my usage, cant live without it.

Im a little jelly of the artifacts feature in claude. Been wanting something like that but their design is better than any of my ideas.

I made a quick bookmarklet too send Github Issue URLs to an api that adds new tasks, or finds extant ones in our internal time tracking utility, and then clocks the team member in on that task. It was afternoon project that the team now uses daily. Never made a bookmarklet before, and my JS knowledge is a WIP, so I used ChatGPT as training wheels.

I built a site that transforms the monthly Who is hiring posts into a structured table [1]. It also allows for semantic sorting to find similar jobs to a given one. It originally used GPT-3, which already worked quite well.

[1] https://www.hacker-jobs.com/

I've built GPTeasers a ChatGPT based quiz generator. You can enter any topic and get a 10 question multiple choice quiz back - which you can play in the web app.

I don't know any webdev so please forgive the jank.

https://djsaunders1997.github.io/GPTeasers/

I built a site where I publish AI generated chat stories, with images and text. The LLM comes up with an idea, plans the story, writes the text messages and generates story images and avatar pictures. Not the biggest success but I learnt a lot!

https://aichatstories.com/

I built an RPG skill tree generator.

You type in a class idea, then you get themed skills back in a skill tree-like diagram, with flavor text for each ability.

Still many things that could be improved, just thought it was fun.

https://rpgskilltreegenerator.com

We have built https://henrik.ai a platform that can build any software you want

A WordPress plugin for WordPress site owners to help them in their day to day queries for managing their website

https://wordpress.org/plugins/quick-assist-ai/

I have a 40 ft HQ container and I would like to fill it with following items:

- 10 boxes of Clothes, 100x140x170 cm, 5kg each, non stackable.

- 8 boxes of Shoes, 80x120x170 cm, 3kg each, keep boxes upright.

- Try to fit everything into a single container.

https://3dpack.ing

I built a text summarizer, key word generator for drilling reports (for co2 sequestration) to try and encapsulate expert knowledge taken when drilling into regression models that predict things like the amount of alterable rock available for co2 storage. It kind of worked.

https://interviewsolver.com

Interview Copilot for helping you ace your live coding interviews. Desktop app + companion web mode so it's truly undetectable by interview platforms, controlled by global hotkeys.

I built a cron job that calls a bash script that gets the current weather forecast for a location and extracts the values and then sshes to another server where it inserts the values into the footer html block of all files ending with .html every hour

I built a client to interface with gguf models or ChatGPT

https://github.com/Merkoba/Meltdown

100% made in Python (and tk). I made my own markdown parser.

Hundreds of commands and command line arguments.

I use it everyday myself.

I built two products out of which one is live - Draftly[1]

We tried different LLM models but always returned back to Open AI.

[1] https://draftly.so

We built a candidate assessment platform. The questions and evaluation is done by LLMs!

Link: https://acciomatrix.com

I built fluent, a chat where all your messages upgraded to the language that you're learning.

https://fluent.im

Local LLM game, guess the secret. Built it using LLama.cpp and whisper.cpp on github.