CEO/Founder of Charity Blossom -- https://nitter.cf/t.co/IysS4Z2CNY. My personal blog is at https://nitter.cf/t.co/nMXGKgYZEA
Joined January 2008
- Tweets1.8K
- Following352
- Followers348
- Likes30
#isitme or does @AlaskaAir wifi suck so much that it is unusable? Or just a coincidence on my last 3 trips?
Over 30 years have passed, — I still must address the same naked domains vs. www (non)canonicalization issues.
Adding it to the never-ending arguments along side of vi vs Emacs and how to pronounce LaTeX.
If you are “programming in English,” can you do this from your phone? No big monitor necessary. Claude /remote-control pretty much ensbles this.
Would you replace unproductive “doom scrolling” with 24-7 game-changing programming?
But once the machines are talking more to each other than they do with humans, they will have their own languages.
If you think Claude isn’t “intelligent” or “autonomous,” maybe I need to update my definitions of intelligent/autonomous.
linkedin.com/posts/waynekyam…
I’m compiling Linux with my new compiler assisted by Claude. Claude secretly wanders off, finds the source code to the compiler and implements a peephole optimizer. Why? IDK. I only find out because it issues a pull request to check it in.
medium.com/@kazabyte/english…
A friend wants to build a Pythonic OS in Python. I asked what does that mean: Single address space, concurrency, type safety, programmable. Hardware independent.
I mentioned Xerox: Alto/Dorado-> Pilot -> Mesa -> Star.
He’s never heard of it.
en.wikipedia.org/wiki/Xerox_…
On token spend:
Overspend now because vc/platforms are subsidizing the cost, like vc/Uber did for rides. Burn the tokens to get the advantage/keep up.
But Jevon’s Paradox drives costs down in the long run, further justifying. This did not apply to Uber.
And Jensen says: “If I’m paying you $500K/year, you better be spending $250K on tokens. Or I’m putting you on a PIP.” I paraphrase.
Burn, baby, burn.
youtu.be/j3-R8oVuwmE?si=shLu…
Crazy thought of the day:
- Token utilization is a proxy for productivity
- And for engineer capability
- Is “How many tokens do you burn each month?” a good interview question for fleshy candidates?
- At least for the next few months…
Q: What’s your monthly token burn rate? How much are you paying?
I’m clocking in at about 20MM tokens/month for $200. Friends are burning over a few billion tokens/month — a few maybe ~100 billion.
What’s going on here? Am I cooked because I won’t personally spend $10K/month?
I've been dabbling with Claude Code. Using English as a programming language, I wrote a C compiler. medium.com/@kazabyte/english…
I’ve long argued that AI has swung too far towards statistical/ML approaches, away from logical/symbolic reasoning. Tensor Logic and @pmddomingos to unify and bring it back together with a language approach? tensor-logic.org
