Infrastructure isn't just about who has the most GPUs anymore—it’s about who has the best compression. The "Middle-Out" era is here.
Don't watch the hype. Watch the weights. 🚀 #AI#TurboQuant#RAMDoubler#TechHistory#MachineLearning
In 1994, doubling your computer's RAM meant either:
1/ Spending $350 on an 8MB hardware stick.
2/ Spending $60 on this floppy disk: RAM Doubler. 💾
30 years later, history is repeating itself. Memory is once again the ultimate bottleneck—and software just broke the wall. 🧵
This is why the "Hardware CAPEX" war is only half the story. Winners? Google and devs running local heavyweights (like my 128GB M4 Max, which now feels like a 700GB powerhouse). Losers? Those betting solely on the scarcity of physical silicon.
@cdecaux Tout a été construit avec Claude Code, l'essentiel des images venant de WikiMedia. La page HTML dans le repo est statique et mise à jour chaque fois qu'on me signale un oubli : Pippin, PowerBook 17 pouces, ...
Builds on @addyosmani's agent-skills (anti-rationalization tables are brilliant). Adds session continuity, grep-able compliance markers, and a tiered knowledge architecture.
The key insight: if a behavioral rule doesn't leave an observable trace, it's decoration.
I kept hitting the same failures with AI coding agents:
- silently guessing instead of asking
- agreeing when they should push back
- forgetting everything between sessions
So here is https://t.co/Rj69KLdGFD
Apache 2.0. Take what's useful.
274 products. From the Apple I ($667 in 1976 = $3,866 today) to the Vision Pro ($3,499).
Every price inflation-adjusted. Bilingual EN/FR. One click to share your total.
Built as a single HTML file. No server, no tracking, no cookies.