@zodchiii What a lie with the recent nerfs that were instituted to the model. It merely dispatches to Opus for real work. What a splendiferous marketing campaign.
"If you don't add AI to your real-life tasks, you will fall like Nokia"
Tell me you've never read the INSEAD case study on Nokia without telling me.
Nokia didn't die because they ignored new tech. They died because fearful leadership killed internal innovation, overestimated their brand, and dismissed software as irrelevant.
"Just slap on AI without understanding it" IS the Nokia mistake.
The only thing dead on that list is the version that never required you to be any good.
Cargo-cult SEO. "Wrap an API, charge $29/mo" SaaS. "Works on my machine" testing. Those are dead. Good riddance.
And if the ground feels like it's moving — it is. That's not a reason to sprint. It's a reason to dig.
"Adapt fast" is just panic with better branding. The thing that actually transfers when the tools change? Understanding. Not speed.
McConkey was told nobody wanted fat skis. Sketched them on a napkin, got laughed out of the room, now they're the entire industry standard. He also raced in a banana suit — because fear of looking stupid was never what slowed anyone down.
Stop writing eulogies for your own career. Go build something that makes the list look ridiculous.
Already building this with Claude Code! My "Jerry" framework:
• Specialized agents (researcher, analyst, architect, critic)
• ps-critic has VETO power (adversarial rival)
• 97 quality gates, 31 checkpoints
• File isolation (no context bleed)
• YAML audit log
2 1 r f ~ N
12-day sessions, 76K msgs
@bcherny Two data points:
1️⃣ Folder that hung: 901 files / 1GB
2️⃣ Current usage: 12-day session still running!
Total: 952 hrs across 87 sessions
Indexing ALL 901 files on every startup = the bottleneck. Happy to share stack dumps or test fixes! 🙏
@bcherny Happy to share:
• Full stack dump (sample output)
• Session file samples (sanitized)
• Repro steps
Love using Claude Code for extended problem solving sessions Would be great to see lazy loading or incremental indexing for the session cache. Keep up the great work! 🙏
@bcherny Data from my ~/.claude/projects:
• 901 session files
• 1GB total (largest: 208MB)
• Date range: Jan 8-31
Workaround: Archive sessions to a backup folder. Empty projects/ = instant startup
The indexing works great for short sessions but doesn't scale long-running ones
@bcherny Found a scaling issue in Claude Code that power users might hit. After 23 days of heavy use (901 sessions, 1GB data), new sessions hang at 99% CPU on startup.
Stack dump shows V8 hot loop in FindOrderedHashMapEntry - session indexing iterates ALL sessions on every launch
What a typical situation of pointing fingers and passing the buck onto someone else. The responsibility starts with parents and continues with the education system. https://t.co/EdODKHnqdh https://t.co/uzMVNXPrlt
Dear @Samsung, the UX of dealing with your "customer support" leaves me in awe. It does not resemble help. Shifting blame to your courier, odd strategy. Cancel + reorder and lose your bonus... cute. That closes your support case but doesn't help me whatsoever. It's a strategy 🤷♂️
Users with iOS13 are about to be rightfully disgusted by how often Facebook spies on their location, so here’s a desperate shit sandwich from the company to excuse the gluttonous tracking.
"Theory is when you know something, but it doesn’t work. Practice is when something works, but you don’t know why. Programmers combine theory and practice: Nothing works and they don’t know why." - Unknown
Some "how harsh has aging hit you" challenge is floating around the internet, which made dig through the archives. On the left is myself circa 2009 vs 2019 on the right. How has the last decade treated you? #agechallenge#agingchallenge#10yearchallenge#10YearsAgoChallenge