kaleb

3.5K posts

kaleb banner
kaleb

kaleb

@KalebAutomates

Data & IT Pipe|Delimiters|Are|Evil| HIPAA Compliant Dennis Nedry saved the world by ruining Jurassic Park Destination: Automation This is the last 30 characters

Tucson Katılım Şubat 2025
80 Takip Edilen442 Takipçiler
kaleb retweetledi
Ale 𝕏
Ale 𝕏@SirAlexthomson·
🚨 SPACE IS LIVE RIGHT NOW We just shipped the article: “They Fired the Team to Save Money Now the AI Bill Is Higher Than the Salaries” Swan AI’s $113k monthly Anthropic bill (4-person team) Uber burning its entire 2026 AI budget in months 55% of companies already regretting AI layoffs Nvidia VP admitting compute costs now exceed employee salaries We’re breaking down the real AI cost hangover, the 55% regret wave, Klarna/IBM rehiring quietly, and what “Talent over Tokens” actually looks like in 2026. Come join and drop your honest take especially if you’ve seen token shock in your company. Space is live 👀 Drop “in” or reply if you want pulled up
Ale 𝕏@SirAlexthomson

twitter.com/i/spaces/1RJjp…

English
0
1
2
113
kaleb
kaleb@KalebAutomates·
@WR4NYGov @elonmusk @GadSaad Watch this person misrepresent my arguments in real time and then call me an intellectual coward
English
0
0
1
9
kaleb
kaleb@KalebAutomates·
@WR4NYGov @elonmusk @GadSaad lmao good one. Called out for selective quoting and you immediately double down with the ad hominem "stop whining." Peak suicidal empathy debate right here 🤡
English
2
0
0
32
kaleb
kaleb@KalebAutomates·
Bad faith is obvious. You selectively quoted half my original sentence to misrepresent it as 'not about the label,' ignored where I explicitly called the behavior zealotry, and now you're moving goalposts with grooming gangs that actually show that exact fanaticism in action. It's ideological capture and political fanaticism, not proven 'suicidal empathy' causation. This is a disingenuous game and doesn't address the actual argument.
English
1
0
0
29
kaleb
kaleb@KalebAutomates·
Selective quoting is a bad look. My original post to Elon was: 'suicidal empathy' is largely a rhetorical label rather than a proven causal mechanism for civilizational self-destruction. The entire sentence is about it being rhetorical rather than proven. That is the point.
English
1
0
1
34
kaleb
kaleb@KalebAutomates·
I've seen the video. It's classic ideological capture, virtue signaling and fanaticism completely overriding self-preservation, like victims who defend their abusers. 'Suicidal empathy' is a useful rhetorical label for when misplaced compassion leads to self-harm. But the root is political zealotry blinding people to tradeoffs and human nature. Simpler than creating a new term, and it explains why her views probably won't change.
English
1
0
0
34
kaleb
kaleb@KalebAutomates·
@WR4NYGov @elonmusk @GadSaad I haven't read the book. My interpretation is based solely on reading the posts about it from the author describing it and others engaging with it.
English
1
0
0
33
kaleb
kaleb@KalebAutomates·
@tszzl its poetic bull shit... literally
English
0
0
0
21
kaleb
kaleb@KalebAutomates·
@Polymarket Because the 'data center' is also a freaking power plant... i swear bro, is no one paying attention?
English
0
0
3
316
Polymarket
Polymarket@Polymarket·
NEW: Residents of rural Texas town complain of “permanent artificial daylight” caused by nearby data centers.
English
182
577
5.9K
448K
kaleb
kaleb@KalebAutomates·
@PeterDiamandis lmfao, yeah a 'commodity' until your AI fails to indent properly and recreates the same script 100 times...
English
0
0
3
44
Peter H. Diamandis, MD
Peter H. Diamandis, MD@PeterDiamandis·
For those worried about any coding language... Your 10 years mastering Python is about to become a commodity. But your ability to think, lead, and create is not.
English
140
97
1.1K
42.8K
kaleb
kaleb@KalebAutomates·
The pattern is: Take ideas that were already well-understood in cognitive science / symbolic AI circles for decades > write a high-profile paper/essay/Substack framing them as the antidote to the current hype > wait for labs to hit the exact walls predicted > claim vindication and "I've been saying this since 2001/2018/2020."
English
1
0
1
41
kaleb
kaleb@KalebAutomates·
@dieworkwear It's so much easier to get lost while reading when the sentences are all clumped together
English
0
0
0
32
derek guy
derek guy@dieworkwear·
Why do people write tweets like this? Where every sentence gets a new line. Sometimes a line might have two sentences. Like this one. But generally speaking, every sentence has a new line, making a tweet look like a long block of text that no one reads. Worse still, such tweets are often repetitive and winding, hammering on the same point over and over again. The writing is often very bad.
English
968
231
7.4K
431.1K
kaleb
kaleb@KalebAutomates·
@GaryMarcus More of a workflow/design problem: no easy way to terminate loops or cap runaway costs. Maybe misaligned by design is better
English
0
0
1
101
Gary Marcus
Gary Marcus@GaryMarcus·
This is actually a version of an alignment problem. Humans have background beliefs (don’t waste large sums of money without telling me) and Claude doesn’t respect those. Caveat emptor.
Om Patel@om_patel5

THIS GUY ACCIDENTALLY BURNED $6,000 IN CLAUDE USAGE OVERNIGHT WITH ONE COMMAND he set a /loop command to check his open PRs every 30 minutes the night before and forgot about it it ran 46 times over 26 hours on opus 4.7. completely unattended here's how claude billing ACTUALLY works: every API call sends your ENTIRE conversation history (not just the latest message). turn 1 sends a few hundred tokens. turn 46 sends 800,000 tokens anthropic uses prompt caching to make this cheaper. if your conversation was sent recently it serves from cache at a 12.5x discount but...cache expires after 5 minutes of inactivity his /loop was set to 30 minutes. so every single iteration: > loop fires, history gets cached > 30 minutes pass, cache expires > loop fires again, cache is gone > has to re-cache the entire conversation from scratch at full price > each iteration also adds its own output so the next re-cache is even bigger by hour 20 the conversation had grown to 800k tokens. every iteration was paying to re-cache 800k tokens at the expensive write rate. the actual PR check responses were nothing compared to the caching cost (obviously) anthropic's dashboard has a multi-day reporting lag. he had no idea anything was wrong until the limit email hit he woke up to a $6,000 bill for a command that was supposed to be a simple PR checker if you use /loop in claude code, keep the interval under 5 minutes so the cache stays warm. or start a fresh session for each loop

English
14
9
105
19.6K
kaleb
kaleb@KalebAutomates·
@OpenAIDevs for the love of god... fix the fucking compact task errors before any more of this fake tamagotchi bull shit
kaleb tweet media
English
1
0
2
326