@ForethinkAI

A privacy-first agent that thinks ahead on your behalf.

forethink.ai/
Joined March 2023
  • Tweets41
  • Following2
  • Followers30
2Media
Forethink retweeted
Adam Benenson
@ashbenenJan 15
Great to be part of this #RustNYC panel — the Rust community keeps growing. I wrote up my full thesis on why Rust's strictness is exactly what makes it easy for AI agents: medium.com/@ashbenen/the-com…

The Compiler Is the Harness: Why Agentic Coding Works So Well in Rust

Rust’s famous strictness — the thing that makes it “hard” — is exactly what makes it easy for AI agents to write.

medium.com
🦀 Rust NYC 🦀@rust_nycJan 14
A phenomenal turnout at today's #RustNYC! Thank you Nominal for being such generous hosts 🤩
2
4
7
592
Digital fragmentation isn't just a UX problem — it's rewiring our cognition. Every context switch taxes our attention, hampering intention. 1/4
1
2
4
218
Replying to @jessmartin @Related
Gotta hear more about your use case, but that’s something we do with @ForethinkAI Can you give an example of the @ tagging use case? Like “@ COMPANY NAME” or “@ PROJECT NAME” type thing?
1
2
90
Replying to @mathemagic1an
You can do this with a few extensions I know of including @ForethinkAI (mine), "Voila", and @getpieces
1
1
4
90
Despite having limited experience with PDF tech, I managed to build a fast, WASM-compatible text extractor by integrating o1 into my Cursor workflow in just 3 hours! The result is efficient and produces code that's bespoke to the needs of my project (clarity vs speed). 🦀⚡️ #RustLang #WebAssembly #AI #PDFExtraction #Programming
1
2
23
5,271
Replying to @jansenniklas
A little bit of both with @ForethinkAI We're building a sort of predictive AI for preparing for tasks and staying on top of multiple projects easily
1
1
63
Replying to @Wattenberger
I want something like this for our browser history timeline at @ForethinkAI As the day moves forward, you get a rear view look at slowly congealing tasks and themes representing the topics and tasks you were engaged with
1
1
3
136
After a month of rebuilding the @ForethinkAI meeting preparation to work more like @cursor_ai, we spent the day refreshing our graphics and copy and preparing a new release (still pending approval)
1
1
2
96
It's never required to continue improving our internal AI-prompt editor, but it's just really fun. Updates from tonight allows us to: 1. preview all placeholders provided from code in context 2. click [⦿] to open your editor to the code (we already have 🤖 buttons to open to the AI execution)
1
3
94
demos for "SoTA RAG" be like
1
1
75
Agreed! This is exactly the conclusion we came to for @ForethinkAI's in-house prompt editor. But, being built into the product means we can easily work with real data while doing our prompt engineering.
Matt Busigin@mbusigin30 Jun 2024
IMO next step for LLM APIs is prompt templating. Claude in particular has clearly realized that meta-prompting is a winning strategy, as well as baking in evaluations into its playground. But the big dog alpha is to do real-time prompt optimization, and this must be done with prompt templates.
1
4
150
We're building an advanced prompt editor directly into @ForethinkAI's dev build. Here's why👇
2
3
9
709
Replying to @labenz @natfriedman
That sounds like the direction of my company @ForethinkAI - browsing history organized by AI and asking questions about recent browsing history (my little work in progress, here)
1
8
494
Load more