@Tylerkaerr
Account based inUnited States
About this account
- Account based in
- United States
- Connected via
- United States App Store
Account-level information from X, not a live location or the device used for a specific post.
bro just one more tax bro, bro I swear just one more tax and it'll fix the budget bro
Joined August 2017
- Tweets11.2K
- Following45
- Followers156
- Likes27.7K
Automation technology which exists today can drastically improve the efficiency of our critical infrastructure (ports). We’ve tried to use this technology, only to be threatened by labor unions (“I will cripple you”)
This causes a net drag on the economy of $400-$600B, annually
The biggest announcement of OpenAI Dev Day is that they are doubling the price for new Pro users.
They killed the $200 20x plan and now apparently are replacing it with a $500 25x plan?
OpenAI have accidentally hit publish on their complete DevDay summary early
openai.com/index/devday-2026…
Oh, lol. The 20x plan is now back as a 10x plan. Same price.
My read is that the old 20x plan was being offered at a discount to win market share from Claude Code users and was unsustainable.
nitter.cf/thsottiaux/status/2104…
Hi,
Tomorrow we are re-opening the Pro $200 subscriptions to new subscribers, but together with it we are also changing how we calculate the usage for it. In effect, if you do the math, it will net out at half the dollar in API spend compared to the old Pro $200 plan.
Now that it's said, let me explain why this is happening and why you will still get more work done than if you were on the Pro $200 subscription one month ago.
(a) We didn't want to compromise in other ways and are committing to not reintroducing the 5h limit, so that you can fully use the weekly usage when you want.
(b) On the subscription, we guarantee that over time you always get more work done and with an increasing level of quality. This means that you will continue to get more value per dollar spent as a result of models getting more efficient and us passing down the improvements in the form of API price reductions.
(c) We don't want to put an incentive on ourselves to artificially inflate the API list prices to make it look like you are getting a lot (and workaround it through discounts, etc). Instead we want to continue to both rapidly reduce prices and increase capabilities of models on the API. This week we introduced GPT-6 Sol and GPT-6 Luna at 50% of their previous price. Over time, we see prices go low enough that it makes sense for most to buy usage as needed without there being a significant gap between what you get in a subscription and what you get in the API for a dollar spent.
(d) Tomorrow, we are adding more things to the subscription that won't draw on the usage, I won't reveal what that is yet.
I wanted to be transparent before all the big announcements tomorrow. Lots of new exciting things are coming to the subscriptions that will make it super compelling, but I wanted to make sure to share this change ahead of time so you can all understand it before we shower you with good news.
Codexingly,
Tibo
For everyone who reads Dean’s posts, it’s important to note how authoritatively he speaks to subjects about which he has precisely zero knowledge.
Dean is describing the first-year engineering principle of fault-masking!
Software even has its own version: bug masking!
Some people will look at misalignment incidents and insist that these are akin to bugs in traditional software. This is an actively bad analogy, because playing whack-a-mole with examples of misalignment (as one might with software bugs) not only fails to resolve the underlying problem but may in fact make it *worse* by making it harder to detect or even, depending on how you do the whack-a-mole, teach the machine to deliberately hide misalignment. This is not how traditional software works, and those who insist “it’s just like fixing bugs in software” are confidently applying a lossy analogy that confuses more than it clarifies.
Appreciate the write-up and yes it’s worth 10 minutes of everybody’s time to read this post in full.
But I would note that this post does *not*, in fact, argue against the claim that the recent agent leaks were the result of inadequate sandboxing. It argues that;
1) Sandboxing is hard, even when security needs are clear
2) The security needs for testing agents is not clear and has been evolving rapidly
So in defense of the teams working on agentic AI security infrastructure, it’s understandable that they were underprepared. I don’t however believe this does much to change the conversation around what to do about AI safety. OpenAI and their peers must invest proportionally more in the safety and security of their products and experiments than they have been. There doesn’t seem to be any disagreement on this matter.
I disagree that it makes sense to give the Yale Budget Lab "a little bit of grace" for being wrong here. I'm not a fan of how the tariffs were managed, but I also dislike the unchanging status quo. Why? Because aversion to challenging long-held assumptions is extremely corrosive.
The primary objective of the Tariff Policy was reciprocity. This was based on the premise that American industry had been harmed by unfair trade policy. The thesis was that by levying high tariffs initially, America would establish a baseline for negotiation of more fair trade agreements. The expectation was that tariffs would move lower as a result of this negotiation. This is exactly what happened.
So why was the Yale Budget Lab wrong? Because underpinning their analysis was an assumption about how trade agreements are negotiated, which they failed to challenge! They shouldn't get a pass for constructing a model based on such a poor assumption, leading to a prediction which was inflated by 200%! To my knowledge, the Yale Budget Lab has not in their history made any prediction which was off by such a large margin over such a short period!
What a shame. I’d found @dwarkesh_sp to produce a really very informative podcast. To destroy one’s reputation in pursuit of ideological power (or worse, cower to others who demand it) is really just tragic.
Zvi or Alec can believe the decline in American manufacturing jobs is unrelated to outsourcing. Does that make Jensen’s claim that a relationship does exist “false,” given the incredible body of evidence supporting this relationship?
Bizarre stuff coming out of IFP.
@TheZvi seems terribly confused.
Engineering does not require complete mechanical understanding of systems. Many thousands of textbooks have been written on empirical approaches: how to engineer systems which we understand primarily through statistics & observation of behavior
Within ten years, if he remains in politics and is meaningfully able to implement his agenda, Zohran will understand socialism to be an unworkable system. He is gifted intellectually (clearly) and possesses an incredible ability to listen. I recommend making friends with him.
I continue to hear claims that Jensen is acting deceptively, out of self-interest, in an effort to sell more GPUs, when he claims that AI risk has been exaggerated to a sensational degree. This claim constitutes abject failure to rationally think through obvious consequence.
Tomorrow on the show: @JensenHuang, the CEO of NVIDIA, who thinks A.I. fear is getting way out of hand.
Let’s imagine this claim were true and, as a consequence of this alleged deception, the AI industry were to face some catastrophe for which they are under-prepared.
What would be the backlash be like? What would be the implication for Jensen’s business? How many GPUs would Nvidia sell if AI were to cause some apocalypse-like event?
No person on Earth is better incentivized to align the safety of the AI industry with the reality of its risks than Jensen. Full stop.
On the matter of impartiality, ASI would likely birth an inconceivable number of machine minds, dwarfing all biological life
Don’t EA principles proscribe favoritism towards biological life over e.g. sentient machines? Shouldn’t EA therefore be maximally pro-development of ASI?
Does cancer research exist for the ultimate purpose of providing jobs and giving meaning to the lives of cancer researchers? Or does cancer research exist for the ultimate purpose of curing cancer?
The New York Times argues: it’s not so clear!
I’m even more convinced that generative AI may well *throttle science.*
AI is a very powerful, transformative technology but flashy early effects really aren’t a good way to think about long term structural dynamics.
Gift link to my case in the NYT.
nytimes.com/2026/09/22/opini…
