where i write: agbocsardi.com
where i ramble: https://www.youtube.com/@agbocsardi
Big on:
📊PhD: networks, strategy, ML
☕️Coffee: yirgacheffe, v60, co-ferment
📸Photos: fuji, edc, analog
Third year PhD, 🇭🇺in🇳🇱
Gergő Bocsárdi
Loading...
New post live: GergőBench, or Benchmarking the Assistant I Actually Use
A small personal benchmark for finding out which AI models can actually handle Cody-shaped assistant work.
Leave any thoughts below!
Applying advanced prompt engineering tactics from a few years ago (which are redundant on frontier models) made a HUGE difference for my experience in getting great work out of open weight models!
I just have a "prompt engineering" skill that Opus or 5.5 uses to write instructions for Deepseek😃
This image is obviously wrong (highest drag should be right in front, not across the top)
And obviously AI
Which I found disappointing
Because instead of just asking ChatGPT or Gemini to make a (wrong) fake picture
You can just ask Codex or CC to rig you up an actual aerodynamics sim
At least it was funny about it😃
shoutout to @lkoro.es for telling me about the github.com/cygnusb/coro...
made me pull the trigger on a pace 4, gonna add some biometric data to my weekly life reviews😃
Got inspired by this post!
github.com/agbocsardi/a...
Super experimental, but hey, it's a start!
someone made a fork of opencode that routes through the unsecured ai endpoints from chipotle
You know what I love about DuckDB? Everything.
final stretch of PhD life: taking ADHD meds and stomach ulcer meds with your 2nd cup of coffee before 9am ツ
Gergő Bocsárdi
Gergő Bocsárdi
Gergő Bocsárdi
Gergő Bocsárdi
Gergő Bocsárdi
Experimental language server diagnostics for academic prose - agbocsardi/academic-lsp
