Sabitlenmiş Tweet
Alex
2.3K posts
Alex
@Willenation
I acquire e-commerce brands, and I’m using AI to automate as much of the business as I can - read about it at https://t.co/ewQMybzR48
San Diego, CA Katılım Ekim 2009
1.6K Takip Edilen357 Takipçiler
@cremieuxrecueil Isn’t this really the original point of venture capital? They funded ventures into the unknown that were highly likely to fail but would bring back tremendous riches if they succeeded.
Doing series E rounds seems more divorced from the genesis of the asset class.
English
@davewingrave @max_spero_ Have you never gotten a note from a human editor and incorporated it because you thought it was an improvement?
English
@max_spero_ Not only that, but you reinforce the idea to yourself that there is an "improvement" that can be provided to you; the doubt is the business model, every use undermines your own agency, you steal from yourself.
English
Honest-seeming post and we shouldn't dunk on it imo My feeling about AI is that you shouldn't use it, ever. My feeling is that of you do, you have to strongly resist the idea that a sentence is a static artifact that can be "improved" outside of its contexts
Shivers@thinkingshivers
I submitted a draft of my short story to Claude for copy editing. Sometimes he’ll suggest a re-write of a particular sentence. My version: “She was just a rich girl, with a carelessness about her that could only come from being born into privilege.” Claude’s suggestion: “She was just a rich girl, careless in the way only privilege allows.” It's a matter of taste, but I personally think Claude's version is better. It's saying the same thing but more deftly. But when I swap his sentence in then plug the paragraph into Pangram, it goes from being high confidence that it's human to low confidence that it's human. If I keep doing this, will it start to read like AI slop? If I keep doing this, is it even my writing anymore? So I'm keeping my version, the one I think is worse, and I'm disquieted by the fact that there could be a better version of this story that I now need to specifically avoid.
English
I don’t think this is always true. I have a pretty distinct sense of what my voice sounds like, and sometimes I can’t get my writing to sound like it (highly correlated with lack of sleep because my kids were up in the middle of the night).
I’ll give Claude sentences that don’t read right or paragraphs where the structure doesn’t work and ask for feedback, and incorporating its notes (which is sometimes taking its rewrite of a sentence, but at the paragraph level usually just rewriting it myself based on its input) gives me a result that better conveys the thought I’m trying to get down on paper.
The obvious parallel here is a human editor. You don’t really hear anyone say that taking feedback from a great editor takes away from the voice of the writer. What’s the difference with using Claude?
English
@davewingrave Every LLM “improvement” erases a small part of the author’s voice and replaces it with the LLM’s preferred voice. Rarely a tradeoff worth taking!
English
As a woman, this is an easy no
What if you happen to be giving birth during those 30 seconds
Your child instantly dies even if you don't
Real Post Folder@RealPostFolder
So are you taking this deal?
English
Today's post is oh-so-cleverly titled The Claude -pocalyspe. Because they're making it more costly to use claude -p. Get it?!
I review my overuse of claude -p for daily jobs (it was effectively free until now, why not) and how I've updated my jobs to mostly stop using it.
English
@Masen711 @njshoreinvest I once read a tweet that said something like, “Being rich means being able to take your kid somewhere on a Tuesday afternoon.” It’s true!
English
@njshoreinvest Being able to be available for your children’s events is the real American dream, you can never get those times back as they grow older 😢
English
The best part about being so close to financial independence is that you stop taking work so seriously.
Instead of logging into our 9am standup call, I’m going to my daughters school for a sing along show they’re putting on.
I didn’t even ask permission. Just pinged my boss on teams and said this is what I’m doing. Don’t care about any repercussions.
Earlier in my career, the anxiety about job performance and stability would have caused me to miss it.
English
@TheStalwart They probably hate AI so much they don’t engage with it at all, can’t recognize it, and thus didn’t have to overcome bias because they had no idea they were looking at AI-generated work.
English
Human judges in a literary competition probably HATE AI.
So think of how good the story must have been in order to overcome that bias.
English
Demand. They’ve been decommissioning old ones thus far because there wouldn’t be enough interest for it to make sense to keep serving them over time. That will change once the models surpass a threshold of utility.
(And yeah I know there was demand for 4o but that was from overly attached people on subscription plans so not economically meaningful. I’m talking about significant usage via API from enterprises who have embedded these models into their workflows.)
English
@Willenation @TheStalwart Why would labs keep the cheaper models hosted and available? They will be rapidly decommissioned
English
So is the correct read on this chart that individual tokens (same model etc) are still getting cheaper. But that as models improve, users are buying COSTLIER tokens in HIGHER volume.
And that this latter effect is moving the PPT needle more than longterm token disinflation?
Liz Thomas@LizThomasStrat
The explosion of agentic AI and compute shortages are pushing up prices: Average LLM token costs are now $2.12/mil tokens,+12% this week alone and +65% since end of Feb.
English
@GarrettPetersen Yeah but if it’s points then it doesn’t count even though I could redeem them for like 75% of the price of the ticket in cash
English
My household earns a lot more than the median, and I can't fathom paying $7500 to sit in a more comfortable chair for several hours. Even if I earned $1M/year, I wouldn't spend 0.75% of that on a few hours in a more comfortable chair.
Scott Hendricks@scottndricks
@ramit @tylerhuff I can easily afford business class.. but never brought myself to do it. Flew with my wife to Bangkok in March and economy was $1,500. Business was $9,000. Can’t justify an extra $15,000 for slightly more comfort.
English
@sporadica These are not the actual values of these positions. Option values are the notional value of the number of shares covered, so the cost he paid to initiate these is somewhere between a little bit and like 100x lower than face value. You have no way of knowing which.
English
what does he know that we dont? bc iirc we are still pretty clearly in a massive GPU shortage, can't we expect semiconductors to keep printing?
Michael Burry Stock Tracker ♟@burrytracker
Breaking: Leopold Aschenbrenner just filed his Q1 2026 13F Here's everything you need to know about his recent 13F Top 10 positions: 1. VanEck Semiconductor ETF $SMH [Put] — $2.04B 2. Nvidia $NVDA [Put] — $1.57B 3. Oracle $ORCL [Put] — $1.07B 4. Broadcom $AVGO [Put] — $1.01B 5. Advanced Micro Devices $AMD [Put] — $969M 6. Bloom Energy $BE — $879M 7. SanDisk $SNDK — $724M 8. Micron $MU [Put] — $584M 9. CoreWeave $CRWV — $556M 10. Taiwan Semiconductor $TSM [Put] — $535M New positions: • $SMH, $NVDA, $ORCL, $AVGO, $AMD, $MU, $TSM, $ASML, $INTC, $GLW — all puts • $MU [Call] — $422M • $TSM [Call] — $355M • $SNDK [Call] — $389M Biggest adds: • CleanSpark $CLSK: +648% shares • Riot Platforms $RIOT: +87% shares Biggest trims: • CoreWeave $CRWV [Call]: -83% shares • Bloom Energy $BE: -36% shares Full exits: • Intel $INTC [Call] — was $747M • Lumentum $LITE — was $479M • EQT Corp $EQT — was $133M • Tower Semiconductor $TSEM — was $85M Summary: He kept his AI infrastructure longs and opened $8.45B in new puts against tech and semiconductor
English
Bench was a nightmare for me as a small business. Egregious mistakes and a constant churn of bookkeepers. Every month they’d say quality check completed and then I’d find some nonsense like huge negative expense numbers that any competent check would’ve questioned. Your product was bad and you did a lot of damage to small businesses with it.
English
Hey @TechCrunch , tell me what you really think. 😂
techcrunch.com/2026/05/14/kho…
English
Asked Claude what my current claude -p usage will run me once they switch to charging API prices for that. This is… slightly higher than my $100/month plan.
To be fair, I really don’t need Opus for any of these jobs. Also I should be able to use scheduled jobs via Claude Desktop or /loop in the CLI.
English
I would check Prinzbench for the best rankings for this use case: github.com/prinz-ai/prinz…
But generally I can say Claude is awful at research and GPT is far better. For very specific stuff like law you probably need particular skills or harnesses - the models can do general reasoning but just don’t have the specific knowledge of how to correctly do legal research. If instructed in detail I suspect they’d get it right.
OpenEvidence has done it well for medicine, but I dunno if anything equivalent exists for law.
Or just wait six months and the core models will probably be able to do it no problem.
English
im doing research for eviction litigation nationwide for different states and tried using AI and it's a complete mess
simple question - is accepting rent during an eviction equals a waiver - every model gives different answers in each state
English