Skip to content

AI

12 products in ai, ranked by the votes people cast.

LangChain

The framework most LLM applications are still assembled with.

AIOpen source

Pinecone

Managed vector search, so retrieval is somebody else's operational problem.

AIFreemium

Ollama

Run language models on your own machine with one command.

AIOpen source

Replicate

Run open-source models behind an API without touching a GPU.

AIPaid

Suno

Full songs (vocals, instruments, structure), from a text prompt.

AIFreemium

ChatGPT

OpenAI's assistant, with browsing, images and code execution built in.

AIFreemium

ElevenLabs

Speech synthesis and voice cloning that survives being listened to closely.

AIFreemium

Hugging Face

The model hub. Weights, datasets and demos for most of open-source ML.

AIFreemium

Midjourney

Image generation with a house style nothing else quite matches.

AIPaid

Perplexity

An answer engine that cites its sources instead of listing ten blue links.

AIFreemium

Claude

Anthropic's assistant, strongest on long documents and careful reasoning.

AIFreemium

Runway

Video generation and editing built for people who already edit video.

AIFreemium

This category covers the tools built directly on large models: assistants you talk to, generators that produce images, audio and video, and the infrastructure the rest of it runs on. They divide roughly three ways. Assistants like ChatGPT, Claude and Perplexity answer questions, with the differences showing up on long inputs and on whether the answer carries citations. Generators like Midjourney, ElevenLabs, Runway and Suno produce media, and each has a house style you either want or work around. Then there is the plumbing: Hugging Face for weights, Replicate and Ollama for running models somewhere, Pinecone for retrieval, LangChain for gluing it together. If you are choosing for a team, the assistant question is usually about data handling and the generator question is usually about commercial rights, neither of which is visible in a demo.

Questions about ai tools

What is the difference between an AI assistant and an answer engine?
An assistant holds a conversation and will happily reason from what it already knows. An answer engine such as Perplexity retrieves sources first and cites them, so you can check the claim. Use an assistant when the task is reasoning or writing, and an answer engine when the task is research you will have to defend later.
Can I use AI-generated images and audio commercially?
It depends on the tool and usually on the plan. Several generators grant commercial rights only on paid tiers, and some restrict them further by output type. Check the vendor's own terms before the work ships, because this is the detail that causes problems later and it is not the one demos highlight.
Do I need a vector database to build with language models?
Only if you are retrieving from your own documents. A vector database such as Pinecone stores embeddings so a model can find relevant passages before answering. If your application only sends prompts and reads replies, you do not need one, and adding it early is a common way to make a simple project complicated.
Is running a model locally worth it?
Locally run models through something like Ollama cost nothing per request and keep data on your machine, which matters for sensitive material. The trade-off is quality and speed: an open model on a laptop is not competitive with a frontier model on an API for hard reasoning. Many teams use both, local for bulk and private work, hosted for anything difficult.
What should a team check before adopting an AI tool?
Whether prompts and uploads are used for training, where data is stored, whether there is an audit trail, and what happens to your content if you leave. Those four answers separate tools that pass a security review from tools that do not, and none of them changes how good the output looks.

Launch your product

Every launch opens Monday. You get a permanent listing, an embeddable badge, and a link you earn.

Submit your product

20 of 20 free slots left this week