
cofounder @cloudcruise_app. building deterministic browser agents
San Francisco, CA
Joined November 2019
- Tweets189
- Following210
- Followers362
- Likes198
Adrian Ziegler retweeted
Chinese bride gifts her partner a flagship GPU at wedding — he breaks down in tears
She handed him an ASUS ROG Astral RTX 5090 that costs around $12,000
Adrian Ziegler retweeted
Opus 5.5 was obviously trained by a much bigger teacher model. Probably Model-2 Mythos. Opus 5.5 is the first model trained from RSI and also distilled from the internal Ant teacher model.
This is how Opus 5.5 is both smaller and cheaper. It is an artifact of teacher model distillation.
Adrian Ziegler retweeted
The degree to which you are awed by AI is perfectly correlated with how much you use AI to code.
Adrian Ziegler retweeted
The Agent axioms:
1. There is always a workflow (it may be trivial such as the implicit single step workflow)
2. Workflows may contain agents (LLMs with tools in a loop)
3. Tools are workflows ((even if trivial per 1) which can contain agents per 2)
Adrian Ziegler retweeted
I've spent hundreds of hours these past few months prompt engineering.
Most of it is tedious trial and error, but I figured I'd share what's worked for us: leveraging markdown vs. xml, how we iterate on inputs, testing different LLMs, etc.
Read more @ cloudcruise.com/blog/prompt-…!
Adrian Ziegler retweeted
There appears to be a massive outage going on...
Twitch, Discord, Google Cloud, Google, Google Meet, Google Nest, CharacterAI, Etsy, Khan Academy, Google Drive, Google Maps, Pokemon TCG, Dialpad, Mailchimp, HighLevel, Amazon Web Services, OpenAI, Cloudflare, Anthropic, Breezeline, Dragon Ball, State Farm, Embark Studios, Gmail, Rocket League, DoorDash, Wells Fargo, Marvel, MLB TV, Google Gemini, Fortnite, Spotify, Shopify, Snapchat, Tekken, Box, Equifax, Roll20, Cursor, Looker Studio, FuboTV, IKEA, reCAPTCHA, GitLab, Steam, Clover POS Systems, AMC Theatres, NPM
Adrian Ziegler retweeted
Replying to @nptacek
This is not the core issue. The core issue is that the LLM has to autonomously and in general way figure out that it is natively not well adapted to do this task in its head, that it doesn’t succeed in doing so, and that it should do this and that instead to solve it.
Adrian Ziegler retweeted
Plz stop making agent demos planning trips or booking flights, how many flights are you even taking??? Make them do repetitive stuff no one wants to do instead
Adrian Ziegler retweeted
With @cloudcruise_app (YC W24), you can build automations that work directly in your browser. Put repetitive work on autopilot like finding sales leads, filling out admin forms, or downloading documents.
ycombinator.com/launches/KbX…
Congrats on the launch @VereWhit, @adrmtu & Felix!
Adrian Ziegler retweeted
I supect "LLMs just predict text" is a Blank Map fallacy. People know nothing else about LLM internals besides that.
Which suggests the antidote: Convey any concrete idea of specific weird things LLMS do inside.
So here's my story about reproducing a weird LLM result...
Adrian Ziegler retweeted
MEGABYTE: Predicting Million-byte Sequences with Multiscale Transformers
abs: arxiv.org/abs/2305.07185
paper page: huggingface.co/papers/2305.0…
Adrian Ziegler retweeted
Today I used GPT-4 to make "Wolverine" - it gives your python scripts regenerative healing abilities!
Run your scripts with it and when they crash, GPT-4 edits them and explains what went wrong. Even if you have many bugs it'll repeatedly rerun until everything is fixed
Adrian Ziegler retweeted
I'm pretty optimistic that the LLM reliability / factualness issue can be fixed. The key is to use LLMs as a dialog interface and not as a store of knowledge. LLMs as the query layer between a human user an a knowledge graph with sources (which can be hybrid generated/curated).




























