Train AI to curate your reading feed
“Prompts can be used to draw images but we wondered: what if we could use them to improve our content feeds?” writes AI maker Philippe Beaudoin, referencing the text-to-image tech you’ve either tried yourself or seen all over Twitter. Beaudoin takes that concept and applies it to your reading feed in his new app, Waverly.
How to customize a feed: With Waverly, you can create your own “waves,” or feeds of articles related to a niche topic that interests you. Or you can follow waves created by others, like “Understanding Consciousness” or “Parenting for New Parents.” From the time you create a new Wave with a prompt (i.e. what you want to read), you’re training the AI to recognize what kind of content you’re interested in. Then you can continue refining the original prompt, or you can do other things that help train the AI, like save articles from the internet into your wave to read later.
The tech: Waverly uses prompt engineering to generate your feed. If you’ve tried ChatGPT or one of the aforementioned image products, you’ve taken part in this concept already. Your input (or prompt) contains instructions or examples that convince large language models (LLM) to produce the content you want. As with any of these products, you might not get exactly what you want out of your first prompt so you might have to refine it, but since prompts are often only a few words or sentences, doing so is quite easy.
Beaudoin, who previously founded ElementAI and was an ex-Googler on the Chrome ML team, has big ambitions along with his co-founders. They believe prompt-driven algorithms that are driven by your intentions and your own plain words will help fight misinformation and echo chambers online.
Is this the game-changer your feed has been waiting for?
Your agents stopped answering and started doing
Your agents call tools, write to databases and move money, so a wrong answer is now a wrong action, which is a considerably worse meeting. Arcjet runs policy before every LLM call, tool call, query and API request: prompt injection, sensitive data, bots, rate limits, spend quotas, and your own rules in Rego on top. It reads the OpenTelemetry you already emit, so switching it on is one environment variable, no code changes and nothing to deploy. Each check also reads the earlier steps in the same workflow, so an agent's tenth action gets judged against the first nine and not on its own. JavaScript, Python and Go SDKs all at 1.0, agent framework guards included, free tier to start on.
Daily Top Products
- Sup lets you play sounds on your friend’s phone… even when it's locked. 😬
- Canua thinks outside the box of single-employer existence and income streams, and helps track all of your financials in one space.
- Seedsummit is a hub for early-stage startups with legal templates and best practices, created in collaboration with law firms and early-stage VCs.
- These ChatGPT prompts can help if you’re still discovering what the tool can do.