Identity != Ideology
Politics are about ideologies. The minute you conflate politics with identity you lose all grounds for productive discussion and debate. Ideas should be judged on the merits of their facts/arguments not on the merits of the person presenting them.
@nateberkopec Been using (and shilling!!) Mise for nearly 3 years now!! At first, I wasn’t sure about leaving rbenv but the fact that I was expanding my toolset and wanted a unified to manage my dev end totally sold me on it…
Mise Tasks is also AWESOME!! 🥰
I think this is true at the founder/founding level: you can have deeper aptitude and talent in one or another area, but AI allows you to accomplish more across the board.
This shift does not scale though… At higher stakes and/or product maturity levels, you need folks who can “hold the ropes” in a specific area, or the building comes crumbling down sooner or later…
not once, not twice, not even just 10 times, `coreaudiod` is the source of absolute chaos on macOS: videos not playing, apps not responding, other apps not opening, etc…
clear symptom this is the issue: "system settings > sound" will not show
and `sudo killall coreaudiod` is not enough
I have to `ps ax | grep coreaudiod`, get the PID, and do a `sudo kill -0 $PID` to kill it and get my system back
GAH
@mitsuhiko There are absolutely 0 reasons to use any other password manager IMO… Been using them since 2008 (beta desktop app on Mac OS X) and they have yet to fail me… which something no other password manager can say! 🫣
So, @okgo just posted a video that surfaces surprisingly good principles about dealing with AI in software engineering… 😂 (highly recommended)
https://t.co/X562ENLP6U
@PeterRHann1 Call it a Tesla and it's fine.
This does not hold Ferrari pedigree…
The interior is gorgeous.
The exterior is a Kia.
The engine is not Ferrari.
Many developers have suspected for months that GPT-5.5 outperforms Claude Sonnet for coding. But SWE-Bench reported near-parity, and it made people question what they’d been seeing in practice.
DeepSWE aligns more closely with that day-to-day experience: GPT-5.5 scores 70% versus Claude Sonnet at 32%. That difference is substantial.
DeepSWE focuses on what tends to matter in real workflows: whether an agent can take a short behavioral prompt, locate the correct area of the codebase, and implement the change cleanly - without needing you to enumerate files, modules, and functions. SWE-Bench often fails to capture that, due to dataset contamination and weaker verification.
https://t.co/C3s80xfDkk
@justinwitz@harish_red@ryancarson Agents can point to their "source of truth". They can present facts and back them up. The issue with agents and hallucinations is when there is no objective factual ground to a decision they need to make… RAG'ing and answering questions are (mostly) solved problems.
I had the same feeling with car racing… I wanted to do more, but the financial demands were just the tip of the iceberg: time, focus, training, socializing, contacts, collaborations, etc, all of that required too much focus that other areas in my life required from me…
It’s hard to stop doing something you love, but sometimes that’s the best decision in order to keep the love for it.