OSS Context-as-a-Service will be huge once we settle on one.
Why is everyone building the same thing?
Vercel indexes its entire GTM memory on turbopuffer https://t.co/vjIw8lIm7E
Another major ✅ Anthropic built C compiler with 16 CC agents consuming $20k of tokens.
What stood out:
1️⃣ Its so true that agents have no sense of time and will happily run useless tests for hours. Feels like something that can be trained into a model like they did with token efficiency.
2️⃣Lots of caveats - just like with the browser project at Cursor - its basically a useless version because of its performance. It took $20k to pass the test harness but it will take 1000x of that to march the 9s.
3️⃣ “We realized we need to break the work down”. Kind of struggle to understand how this keeps being rediscovered over and over again. I mean - how do you eat an elephant? With a spoon.
Nevertheless - its SO cool!
https://t.co/P7mRZLltJX
Cockroach score: simplest way to make sense of SaaSpocalypse.
🪳score = Hours spent in UI / Avg revenue per user.
Low-ARPU/high-use software is undisruptable. It does not leave room for underserved users. Especially not with subsidized $$$ tokens.
Chart below ranks by this relative to 52w drawdown and flags FCF/SBC/buyback stuff.
🦀 Everything is crab again 🦀
Flood of ambient CRMs incoming.
“Every program attempts to expand until it can read mail. Those programs which cannot so expand are replaced by ones which can.”
🦀 Everything is crab 🦀
“Every program attempts to expand until it can read mail. Those programs which cannot so expand are replaced by ones which can.”
Codex = Cursor = Antigravity = Conductor
All AI coding apps have evolved into an inbox.
P.S. TIL that nature has evolved crabs multiple times: https://t.co/ahrcqUypuj. Crabs dominate.
@wesmckinn@wesmckinn brilliant. Thoughts for attachment pipelines. I have file->.md CLI and qmd. Whats best to turn my gmail into searchable markdown?
qmd on CPU and portable index please.
I have a PR that:
- adds “remote” cmd to be able to hit ollama or litellm remotes for the 3 models used
- adds “init” (or —index-path) to remember where index sits (helps when u want to index once on mounted drive and call qmd anywhere)
do u accept PRs? Or you can just prompt - these are simple.
@DhravyaShah Error searching memories: 403 {"error":"The Clawdbot plugin requires a Pro plan or higher. Upgrade at https://t.co/m8xMUv1JoQ"}
That's lame - cant even try it
@schickling Thats what an IDE already is. Its an Interactive Development Environment, not Improved Dev Editor. Most people just dont know how to use it and given more blank slate textbox kniw even less than extension UIs.
@jerryjliu0
llama index is great - been using it since 2024.
But you are really missing out on an opportunity to move up from OCR to Context-as-a-Service:
* Embrace "all-agent-needs-is-a-folder" - along the lines of what you have been posting re: CC + llama CLI
* Create OneDrive/Box/Dropbox connectors people can use with delegated OAuth tokens to turn shares into parsed folders
* Create Gmail/O365 email to turn mailboxes into parsed folders (w/ parsed attachments)
Make this available as an SDK where people can compose higher-level agents (v0, shortcut, cc) on top with some basic tools like LanceDB.
If something like this existed without becoming Glean or becoming Langchain - there is an untold amount of demand in SMB/verticals for it. Everyone is just re-implementing harnesses and file syncs this year.
Make Claude Cowork actually understand complex research reports 📊📑
Claude Cowork is awesome, and it’s great at reading plaintext files, but it lacks the ability to accurately/efficiently analyze complex document formats, like PDF, Word, Pptx, etc. (more on this in a thread later; tl;dr it needs to write too much stuff from scratch).
I made a simple command that can export any complex document into high-quality markdown, courtesy of LlamaParse. In the video below, we parse all visuals in this CB Insights report into markdown-formatted text and tables.
This lets Claude Cowork do analysis over any document and give highly accurate responses in seconds.
My simple wrapper over semtools to batch export files into markdown: https://t.co/Et8IwvxHsY
Sign up to LlamaCloud: https://t.co/qsGFaIL7uM