engineer @braintrust, prev @supabase @sentry
Vienna, Austria
Joined July 2013
- Tweets809
- Following466
- Followers288
- Likes2.3K
idk man this doesn't seem like a good experience. Why did it ask back? it could have just repro'd it itself. put your stuff into a sandbox and give it more powers. This is 50% wasted interaction.
It's 8am on a Sunday, I notice the @linear desktop app tab bar doesn't allow window dragging anymore.
File a bug to triage, Linear agent asks me some clarifications, we find out it's related to scroll position. Linear starts a coding session to make a fix. PR opened with AI code review guide to explain the changes
Seamless
my first Jev use-case: show a model suggestion in my coding ui based on the prompt. pressing tab accepts the model suggestion. which model to use is configured with natural language and takes previous turns as context so you can write things like "no do it differently please"
So we have like half a year before the singularity? kk
AI systems are getting more powerful, and they're increasingly being used to build the next version of themselves. We want to illuminate that progress for the public.
Today, we're sharing three measurements that help track AI development:
1. How much AI R&D is done by AI.
2. How well AI agents are overseen.
3. How compute is allocated.
We provide a snapshot of these metrics from inside Anthropic. Any frontier developer could publish the same measures, and third parties could verify them.
As the world considers pacing the frontier, we should do everything possible to minimize the gap between what frontier labs know and what the public knows. This means better measuring the development of AI, publishing our findings, and giving society an opportunity to decide how to use this information.
Read the full post and methodology: anthropic.com/institute/meas…
I am starting a petition to stop extending promises again github.com/typesafe-ai/types…
@typesafeai you guys can fix this before you cause pain for your users for years to come. don't follow the O's and A's of your industry here pls
Told codex to summarize my communication style on github so I can tell it to write like me: gist.github.com/lforst/43d0d…
fun exercise, will report back whether it works in terms of producing less annoying text
prompting my agent to regularly prompt itself to prompt the coding agent to improve the agent prompt
Luca Forstner retweeted
You could move to Austria
Make €2,700 as a PhD researcher for the next 40 years
Live in this beautiful flat for €900 a month
Retire comfortably at age 55 with full government benefits
Never worry about healthcare costs
Drink a large mug of beer every day at 15:00 when work ends
What are you waiting for?
I got so used to GitHub being janky and slow that whenever they change something not to be janky and slow, and I interact with it, I immediatly think that something can't be right or trusted.
Free billion dollar idea up for grabs: Grammarly for prompting in coding agents. Next time you're about to write something ambiguous that the agent will gloriously mess up - have clippy pop up and warn you.