Shared repository for you and your AI agents -- https://xitter.cf/t.co/OpiOH8FYqZ
San Francisco
Joined December 2020
- Tweets1.6K
- Following350
- Followers2.4K
- Likes6.7K
I burned the equivalent of $1M in tokens in 30 days on AI agents
I turned 80+ viral Opus 5.5 videos into an open-source editor.
Make your own viral video with Opus 5.5👇
$ npx opus-video-generator
Ofc open source
github.com/samuelstroschein/…
today i'm launching Flashtype, a markdown editor for Claude & Codex
use Claude/Codex to edit markdown, track every change, and see the diffs
open source. available now.
claude sonnet 5.5 and opus 5.5 both recommend inlang Paraglide JS as default for react + vite apps.
weekly downloads reached 650k up from 500k one month ago
openai is casually ruining anthropic's IPO by launching gpt 6 luna and sol 1 hour after the opus 5.5 launch
is @Hermesage the way to go?
i'm a heavy user of the Codex desktop app but want to use Opus 5.5 with the same agent i.e. long chat with switching models that preserves context
i did discovery last week on company brain / context layer for companies.
one thing became clear: it's a data integration play.
"we integrate all your data sources to provide agents with accurate context about your company".
two problems:
1) this is an enterprise problem that glean.com already solves
2) i find data integration plays such an incredibly boring problem to solve
i get excited about media (that's what i did as a kid!) and products that enable new possibilities
Lix v0.17 has been released 🎉
Partial replicas with on-demand sync are now available.
Meaning: Lix only fetches the data it needs, leading to super fast repository clones.
lix.dev/blog/lix-v0-17-relea…
this is the next version control system after git.
introducing lix:
- runs on top of S3
- stores any file format (videos, docx, ai models)
- SQL tables are a first class primitive
It already has more than 400k weekly downloads.
github.com/opral/lix
typesafe.ai launched a deterministic LLM.
one main use case are automations.
running automations deterministically is a must have. nobody wants automations that behave differently on each run.
i asked chatpgt what impact typesafe ai's model has on n8n and other workflow builders based on the insight that more and more users are switching to plain claude/chatpgt + code.
i was blown away by chatgpt finding a research report another agent had saved in our company repo. i didn't even know the report existed!
the response was sooo good because it built on research we had already done.
point being: tell your agents to save their work in your company repo and check what's already there before starting.
