A compute system targeted at making people embrace technology.
Other processes include building Minecraft contraptions and questioning Standard Particle model🥑
@jeffplatt You being hard on yourself.
Interviewing someone right after a bust is a tricky situation.
And a legend like Phil Ivey in that zone must be intimidating !!
I'm tired of getting 3 likes man
I want brothers(and sisters) in new tech, AI, startups, product, distribution, vibecoding, nyc or elsewhere to build with on Twitter with.
@deadpixeldavid@sickdotdev The abstraction done for computer use and how to pipe to leverage bash/system resources is added overhead.
Both in context size (extra injected init prompt) as well as per turn usage
@sickdotdev For me personally it computer use.
To be more precise the transparent computer use.
The power is not just scripts or code - but to actually run it, iterate on it.
We can do this with GUIs as well, however there is thin abstraction done there making few ops opaque
Currently renting in Bellendur area.
I took a place close by to office thinking i would be able to skip rush by walking or cycling.
After an year or so. I stopped walking completely- NOT safe!
Now takes me about 20 mins to reach office- less than a KM - contributing to the daily ruckus….
Sad state of affairs
I am pretty sure this is not new and it is not just fable.
Above the harness - you could ‘feel’ that opus injects like a “delight” prompt.
Sonnet class models has that inherent DAG before executing.
- as soon as we go for “thinking” modes for the models - that means pushing for some agent loops inside the /completions API
@mitsuhiko Hahaha!
I choose to believe for the next few hours that this is true - that a suggestive/opinionated piece on a personal git project started a nudge globally to go sleep!
Love this
@mattpocockuk How we limit the multi-step part.
I do gravitate towards that sometimes.
Encoding SOPs ,behaviour characteristics.
Also, harness agnostic is good in theory. But as agent runtimes are different - this part also gets encoded or referenced
Still figuring out the right ‘human in the loop’ setup with agents. Would love your thoughts or suggestions.
Try this template for your own projects: https://t.co/GyoEYleesm #PKM
I’ve been experimenting with ‘gardening’ my notes instead of just collecting them.
Modified PARA into Pipeline, SkillSurface & Crystallized notes + maturity frontmatter.
Surprisingly, it works really well even for project & product notes, not just a full second brain.
Bengaluru, Rasthe is live 🔥
Now better than ever. We took your feedback and made the app even better with @10x_apps
This version now holds the authorities and BBMP accountable
Try it now : https://t.co/okMDtQ7GXE
The mobile app version is up for review on the AppStore, the web version is functional now.
Thank you for all the support
Let’s make Bengaluru better.
I fixed Bengaluru and TRAFFIC in 30 minutes.
Meet RASTHE: Tinder for Footpaths
A crowdsourced platform where you can:
• Report broken / missing footpaths directly to BBMP
• Swipe footpaths like Tinder
• Upvote the worst spots that need urgent fixing
• Turn every citizen into a sensor for the city
If roads are built for cars, this is built for people. prevents pedestrians from walking on roads reducing congestion
Built and ready to submit to AppStore using @10x_apps in 30minutes only
How can we get OpenCode and Pi on these benchmarks?
I know these benchmarks are not cheap to run - but for the sake of growing the open tools, even a community crowdfund to make it happen would be worthwhile (I know they are both backed by for profit companies now, but tiny compared to the BigLabs and Cursor who can afford to run the benchmarks on every 0.1 release)