Peter Herzog
4.7K posts

Sabitlenmiş Tweet

Wide release of FSD 14.3 in a few weeks 😱😱

Elon Musk@elonmusk
@DBurkland @pbeisel It’s in testing right now. Wide release in a few weeks.
English

@dwr Wispr Flow or Typeless or Monologue. They are way better than anything else. It's hard to tell the difference between these three. They are stunningly good.
English

😳
Jonathan Haidt@JonHaidt
More evidence that the global decline in test scores that began after 2012 is linked to the proliferation of smartphones and computers in class: The slide was bigger in countries where students began spending more time on devices (for leisure) generationtechblog.com/p/phones-at-sc…
ART



The U.S. Federal Aviation Administration (FAA) issued a Notice to Airmen (NOTAM) Wednesday morning entirely closing the airspace over the Western Texas city of El Paso for the next ten days, citing “special security reasons,” while classifying a large area over both El Paso and the neighboring community of Santa Teresa, New Mexico as “national defense airspace” under the control of the Defense Department. All flights in and out of El Paso International Airport and nearby regional airports have been canceled or grounded, with even medical evacuation flights being restricted, as local authorities and military officials state that they have received no information regarding the NOTAM. Such an airspace closure over a major city in the United States is unprecedented, with the last time such a restriction like this was issued occurring after September 11, 2001.

English
Peter Herzog retweetledi

First day with FSD v14.2.2.3. Lots of driving today. Lots of parking. Parking seems improved! Very happy with choices!
On the other hand, observed two occasions where the car rapidly turned left and right, as if it couldn’t decide the best direction. Once when reversing from a parking spot and once when turning left onto another road. Both times going pretty slow, thankfully! @Tesla_AI
English
Peter Herzog retweetledi

@steipete How often do you create new codex session/thead/window to get a fresh context window? Do you have an automation or process that passes context from session to session? Or do you just rely on the compaction?
English

My favorite way to refactor and keep the codebase fresh: I use one context.
I queue a lot of little tasks, and at some point, I ask the model what it would like to refactor - by then the model has probably seen a lot of code and "felt the pain" and will have quite good ideas on how to restructure the codebase.
I used to create more separate contexts, but with GPT 5.2, it's no longer necessary. It has extremely good performance even when the context is almost empty, so I could drop this charade.

English

Seahawks are number 1 seed in NFC.
Starting QB: Sam Darnold.
Darnold as a Viking in 2024:
He started all 17 games, leading the Vikings to a 14-3 record and a playoff berth. He went to the pro bowl.
His key stats were:
• 361 completions on 545 attempts (66.2% completion rate)
• 4,319 passing yards
• 35 touchdown passes
• 12 interceptions
• Passer rating of 102.5
2025 season, Darnold’s stats:
• Approximately 285 completions on 425 attempts (67.2% completion rate, estimated from yards and average)
• 3,703 passing yards
• 24 touchdown passes
• 13 interceptions
• Passer rating of approximately 100.6
English








