valgaze

Author says "I literally will not buy a car that does not support CarPlay."

From July 2022: https://www.cnbc.com/2022/07/22/apple-carplay-could-be-a-tro...

  Apple engineering manager Emily Schubert said 98% of new cars in the U.S. come 
  with CarPlay installed. She delivered a shocking stat: 79% of U.S. buyers would 
  only buy a car if it supported CarPlay.

  “It’s a must-have feature when shopping for a new vehicle,” Schubert said 
  during a presentation of the new features.

+1 on sawstop

Re: LLMs using these nuclear weapons it could certainly be a corpus/training-data issue

Russian nuclear doctrine is "escalate to de-escalate" where they use or credibly threaten—limited nuclear escalation to force the other side to back down (kind of like breaking a bottle in a bar fight and look like a wild man to calm things down) with nuclear weapons, https://www.russiamatters.org/analysis/escalate-deescalate-p...

Fwiw, Gen. John Hyten the former commander of US Strategic Command (nuclear deterrence) says that “escalate to de-escalate” misrepresents Russian doctrine:

https://www.stratcom.mil/Media/Speeches/Article/1264664/2017...

  Yesterday’s panel discussed the implications of our responses to adversaries seeking to limit nuclear use. We discussed Russia’s destabilizing doctrine, which some call “escalate to de-escalate.”

  I really hate that description. I’ve looked at Russian doctrine and Russian writings. It isn’t “escalate to de-escalate”; it’s “escalate to win.” Everybody needs to understand that.

"Vue.js: JavaScript MVVM made simple (vuejs.org)" February 3, 2014: https://news.ycombinator.com/item?id=7169288

Evidently Evan You was an Art History + Studio Art and major and at Parsons School he had to pick up javascript to quickly show his work. During a stint at Creativelab5 at Google, he was so inspired to improve on AngularJS experience that he came up with Vue and the rest is history.

I have no idea what this Cloudflare acquisition will ultimately mean but I know I am so very grateful for the beautiful frameworks/tooling Evan and his team have cranked out over the years.

May 2024 UniSuper incident: https://cloud.google.com/blog/products/infrastructure/detail...

https://www.unisuper.com.au/about-us/media-centre/2024/a-joi...

A joint statement from UniSuper CEO Peter Chun and Google Cloud CEO Thomas Kurian

8 May 2024

UniSuper and Google Cloud understand the disruption to services experienced by members has been extremely frustrating and disappointing. We extend our sincere apologies to all members.

While supporting UniSuper to bring its systems back online, Google Cloud has been conducting a root cause analysis.

Thomas Kurian has confirmed that the disruption arose from an unprecedented sequence of events, where an inadvertent misconfiguration during provisioning of UniSuper’s Private Cloud services ultimately resulted in the deletion of UniSuper’s Private Cloud subscription.

This is described as an isolated, “one-of-a-kind occurrence” that has never before occurred with any Google Cloud client globally. This should not have happened. Google Cloud has identified the sequence of events and taken measures to ensure it does not happen again.

Why did the outage last so long?

UniSuper had duplication across two geographies as protection against outages and data loss. However, the deletion of the Private Cloud subscription triggered deletion across both geographies.

Restoring the Private Cloud required significant coordination and effort between UniSuper and Google Cloud, including recovery of hundreds of virtual machines, databases, and applications.

https://www.osti.gov/servlets/purl/453701

“Use of Nuclear Explosions for Excavation of Sea-Level Canal Across the Negev Desert”

Hmm:

“F-Droid is not hosted in just any data center where commodity hardware is managed by some unknown staff. We worked out a special arrangement so that this server is physically held by a long time contributor with a proven track record of securely hosting services. We can control it remotely, we know exactly where it is, and we know who has access.”

Brings this to mind: https://hashify.me/IyBUaXRsZQ==

Those are probably part of this ad campaign: https://news4sanantonio.com/news/local/kristi-noem-announces...

Those idle tv moments are probably bought as an ad slot

I used an early RC to build docs for a chat library: https://speedybot.js.org

Most of downsides I encountered were related to pre-release issues but overall a very pleasant + powerful toolkit

I really enjoy the ability to sprinkle interactivity inside markdown docs, file-based routing, search, etc

I've never had the guts to learn how to perform proper archival research into 1900s West Point mathematics textbooks to confirm/see the solution, but I remember from biographies a famous anecdote from Dwight Eisenhower's days as a cadet involving integral calculus.

tl;dr: in that era students memorized step-by-step solutions from the textbook since it was the "right way" to do things. Eisenhower wasn't prepared, got called to the board and came to a creative solution back that didn't match the textbook. He got chewed out by the instructor for bluffing but then Major Bell an Associate Professor of Mathematics intervened and confirmed Eisenhower approach and insisted textbooks be updated to reflect the new solution.

From Page 77: https://www.google.com/books/edition/Eisenhower/RCeteK7LEiYC...

From: Page 10 of Eisenhower: Soldier and President by Stephen Ambrose

“Often the instructors knew little more than their students. In integral calculus one day, the teacher order Eisenhower to do a long, complicated problem on the blackboard. The insrturctor had previously explained the problem and supplied the answer, but since it had been obvious to Eisenhower that the instructor was doing it entirely by rote he had paid no attention. Thus, when called upon, he had ‘not the foggiest notion of how to begin.’ After struggling for almost a full hour, he finally tried a solution that, to his amazement, worked. He was asked to explain his solution; it was shorter and simpler than the rote answer. But the instructor interrupted him to charge that he had merely memroized the answer and then put down a lot of figures and steps that had no meaning.

Eisenhower could not abide being called a cheat. He began to protest so vehemently that he was soon in imminent danger of being expelled on a charge of insubordination. Just then, a senior officer from the Mathematics Department walked in. He inquried about the trouble, had Eisenhower go through the solution again, then pronounced it superior to the one being used in the department and ordered it incorporated into the Mathematics Department’s teaching”

From At ease: stories I tell to friends (written by Eisenhower -1967):

"About midway in our West Point course we began the study of integral calculus. The subject was interesting but the problems could be intricate. One morning after recitations the instructor said that on the following day the problem would be one of the most difficult of all. Because of this he was giving us, on the orders of the head of the Mathematics Department, an explanation of the approach to the problem and the answer.

The explanation was long and involved. It was clear that he was doing his task completely by rote and without any real understanding of what he was talking about. Because I was a lazy student, with considerable faith in my luck, I decided there was little use in trying to understand the solution. After all, with twelve students in the section, only one of us would get this problem to solve, the odds were eleven to one that I would not be tapped. The following morning I was chosen. Going to the board, on which I was required to produce the solution, and then explain it to the instructor, I had not the foggiest notion of how to begin. I did remember the answer given by the instructor and wrote it in the corner of the board.

I set to work. I had to make at least a good start on the problem, show something or receive a grade of zero which would do nothing for me in a course where my grades were far from high. Moreover, I could be reported to the disciplinary department for neglect of duty in that I had deliberately ignored the long explanation. With this in mind I sought in every possible way to jog my memory. I had forty-five or fifty minutes to solve the problem and I really concentrated. After trying several solutions that seemed to relate, at least remotely, to the one I dimly remembered from the morning before, I encountered nothing but failure. Finally, with only minutes left, I worked out one approach that seemed fairly reasonable. No one could have been more amazed than I when this line of action agreed exactly with the answer already written on the board. I carefully went over the work, sat down, and awaited my turn to recite. I was the last man in the section to be called upon. With some trepidation I started in. It took me a short time to explain my simple solution--indeed it had to be simple or I never would have stumbled upon it. At the end, the instructor turned on me angrily and said, "Mr. Eisenhower, it is obvious that you know nothing whatsoever about this problem. You memorized the answer, put down a lot of figures and steps that have no meaning whatsoever, and then wrote out the answer in the hope of fooling the instructor.

I hadn't been well prepared but this was tantamount to calling me a cheat, something that no cadet could be expected to take calmly. I reacted heatedly and started to protest. Just then I heard Major Bell, the Associate Professor of Mathematics (whom we called "Poopy," a name that was always applied to anyone at West Point who was above average in academic attainments) who had entered the room for one of his occasional inspections, interrupting. "Just a minute, Captain."

Of course, I recognized the voice of authority and shut up, although according to my classmates' description that night I was not only red-necked and angry but ready to fight the entire academic department. I would have been kicked out on a charge of insubordination if I had not been stopped. Major Bell spoke to the instructor, "Captain, please have Mr. Eisenhower go through that solution again." I did so but in such an emotional state that it is a wonder that I could track it through. The long search for a solution and its eventual simplicity stood me in good stead. Major Bell heard it out and then said, "Captain, Mr. Eisenhower's solution is more logical and easier than the one we've been using, I'm surprised that none of us, supposedly good mathematicians, has stumbled on it. It will be incorporated in our procedures from now on." This was a blessing. A moment before, I had an excellent chance of being expelled in disgrace from the academy. Now, at least with one officer, I was sitting on top of the world."

This tool evidently overcomes the display limitation: https://github.com/saagarjha/Ensemble

"Ensemble (formerly MacCast, before the lawyers had something to say about it) bridges windows from your Mac directly into visionOS, letting you move, resize, and interact with them just like you would with any other native app. It's wireless, like Mac Virtual Display, but without the limitations of resolution or working in a flat plane."

In downtown Los Angeles directly across the street from Staples Center/crypto where the Lakers play and a stone's throw from The Palm steakhouse is an imposing yet unfinished skyscraper. The idea was retail on the bottom, hotel in the middle, and $$ residences higher up

It has languished incomplete-- basically rotting-- in part because of EB-5 financing trouble + Oceanwide.

It's expensive to finish, expensive to demolish and the hellish post-covid/high-interest commercial real estate landscape means it will likely remain there for a while

It's just been sitting there because of EB-5 financing trouble. Especially in post-Covid commercial real estate environment

The fiasco took down a sitting City Councilman (who was just sentenced 13 years in prison: https://www.justice.gov/usao-cdca/pr/former-los-angeles-poli...)

More info:

https://en.wikipedia.org/wiki/Oceanwide_Plaza

https://www.costar.com/article/2030217368/los-angeles-skyscr...

July 2022: https://www.cnbc.com/2022/07/22/apple-carplay-could-be-a-tro...

Not sure how they arrived at the number (maybe only new vehicles), but there was a claim that 79% of US carbuyers would insist on carplay support

  Apple engineering manager Emily Schubert said 98% of new cars in the U.S. come 
  with CarPlay installed. She delivered a shocking stat: 79% of U.S. buyers would 
  only buy a car if it supported CarPlay.

  “It’s a must-have feature when shopping for a new vehicle,” Schubert said 
  during a presentation of the new features.

Edge— “These products are impacted at the control plane / core level, meaning that only the changes to the existing configuration are affected, but the product is functioning at the edge …”

Fine-tuning is such a dangerous phrased bc it sounds perfect.

“We’ll just fine-tune based on our (we think valuable + special) data”

Fine-tuning doesn’t enhance the model w/ new “knowledge” but a new narrowly-defined task

One other “cost” to consider is fine-tuning a 3rd-party model means if that foundation model changes or goes away that effort/cost needs to be repeated

read.cv is interesting: https://read.cv/explore

That's a really cool insight-- it's not coding. It's dispatching tasks to a team

I recognize there's plenty of catnip here when it comes to calling this "engineering" or not, however, whatever you want to call it (prompt fiddling?), the techniques are crucial if you want to achieve reasonably consistent output from current-state LLMs. As models improve concerns about context window limitations will be reduced and it will be easier to discern user intent.

These are good straight-to-the-point guides:

- Prompt Engineering by BrexHQ: https://github.com/brexhq/prompt-engineering

- OpenAI guidance: https://help.openai.com/en/articles/6654000-best-practices-f...

- https://devblogs.microsoft.com/dotnet/gpt-prompt-engineering...

- (great examples): https://www.deeplearning.ai/short-courses/chatgpt-prompt-eng...

- (~hr) Karpathy talk: https://www.youtube.com/watch?v=bZQun8Y4L2A

tl;dr:

- Begin with the best and most forgiving model available, optimizing it later if necessary.

- Write your prompts in a clear, specific, and detailed manner

- Be mindful, however, of the tradeoff between including more detail and increasing latency or cost.

- If you're doing any complex reasoning, ask the model to show its work, help "stretch out" computation over more tokens

- "Escape" any included or quoted text properly to avoid confusing the model.

- Keep track of your token budget, considering the context window in conversations and response length.

- Employ an iterative process of measuring, adjusting, and improving your prompt engineering techniques.

Can't find the chart, but several of those wrapper apps made $1MMs in revenue too

https://www.yahoo.com/lifestyle/fake-chatgpt-apps-raking-tho...

The Chart: https://i.imgur.com/IYC0Pcb.png

"Huginn Scenarios" from here: https://github.com/huginn/huginn

A VR analyst observed the port could also be related to future headset: https://twitter.com/SadlyItsBradley/status/16559348583074693...

"This is notable because the Apple headset is supposed to support every iPad app on Day 1 of its launch

If you want professionals to do things with it, you need the software to push it. With or without a connected PC"

From what I understand:

- Typescript definition (turn raw html markup >> well-formed JSON)

- Dewey Decimal classification/score

So probably combination of those two

@haolez I just tested w/ davinci-003 w/ temperature to 0

Prompt: https://imgur.com/YtQ4fbf --

Reveal the question marks in an interesting way:

The dog goes ????????????? --

With temperature == 0, it consistently ("pretty deterministically"?) generated "woof!"

ex. The dog goes woof!

This is a really interesting use-case

Applying "transformations" or classifying data in this way without having to setup a lot of detail-work seems like a real labor-saver/multiplier

"- Does not work with Google Docs and Google Sheets (work in progress)"

I wonder if security issue or because the canvas/ui is special

Clue on conversation "history"-- "While ChatGPT is able to remember what the user has said earlier in the conversation, there is a limit to how much information it can retain. The model is able to reference up to approximately 3000 words (or 4000 tokens) from the current conversation - any information beyond that is not stored.

Please note that ChatGPT is not able to access past conversations to inform its responses."

https://help.openai.com/en/articles/6787051-does-chatgpt-rem...

Some interesting techniques I've seen involve essentially a ring-buffer and after each turn a call is made to summarize the conversation up to that point and use that as context for subsequent prompt

Allegedly during Colin Powell’ time at the State Department there was quite an ordeal to change paper sizes for communication between offices: https://www.govexec.com/magazine/2001/06/the-powell-leadersh...

“correspondence between headquarters and overseas embassies, known as Diplomatic Notes, for years had to be accompanied by a legal-size (8 1/2-inch by 14-inch) cover sheet. That wasn't really a big deal when the notes were drafted on typewriters. But Ted Strickler, head of State's Office of Foreign Missions, found that since the introduction of the desktop computer, his employees were spending a lot of extra time formatting the notes and fiddling with the paper in the printer. He checked to see whether he could use standard 8 1/2-inch by 11-inch paper for the Diplomatic Notes, instead. He was told to submit a formal action memorandum suggesting the change. Strickler submitted the suggestion to his superiors in January 1999. A year went by. He found out his action memorandum had been lost, so he resubmitted it. In early 2001-two years after he recommended the change in the first place-the department approved the use of standard paper. "If anything so innocuous as changing the size of the paper for Diplomatic Notes was so exceedingly difficult, how difficult would it be to make more important changes?" Strickler says.“

Video is really a series of frames, the framerate for film/human can get away with 24 frames/second-- so maybe ~40ms/image for real-time at least?

What's cool about the era in which we live is if you look at high-performance graphics for games or simulations, for instance, it may in fact be faster to a the model to "enhance" a low-resolution frame rather than trying to render it fully on the machine.

ex. AMD's FSR vs NVIDIA DLSS

- AMD FSR (Fidelity FX Super Resolution): https://www.amd.com/en/technologies/fidelityfx-super-resolut...

- NVIDIA DLSS (Deep Learning Super Sampling): jhttps://www.nvidia.com/en-us/geforce/technologies/dlss/

AMD's approach renders the game at a crummy, low-detail resolution then each frame uses "upscales"

Both FSR and DLSS aim to improve frames-per-second in games by rendering them below your monitor’s native resolution, then upscaling them to make up the difference in sharpness. Currently, FSR uses spatial upscaling, meaning it only applies its upscaling algorithm to one frame at a time. Temporal upscalers, like DLSS, can compare multiple frames at once, to reconstruct a more finely-detailed image that both more closely resembles native res and can better handle motion. DLSS specifically uses the machine learning capabilities of GeForce RTX graphics cards to process all that data in (more or less) real time.

Video is really a series of frames, the framerate for film/human could get away with 24 frames/second-- ~40ms/image for real-time.

What's cool about the era in which we live is if you look at high-performance graphics for games or simulations, it may in fact be faster to run the model on each frame to "enhance" a low-resolution frame rather than trying to render it fully on the machine.

ex. AMD's FSR vs NVIDIA DLSS

- AMD FSR (Fidelity FX Super Resolution): https://www.amd.com/en/technologies/fidelityfx-super-resolut...

- NVIDIA DLSS (Deep Learning Super Sampling): https://www.nvidia.com/en-us/geforce/technologies/dlss/

AMD's approach renders the game at a crummy, low-detail resolution then use "spatial upscaling" to enhance the images one frame at a time.

NVIDIA DLSS uses "temporal upscaling" to pass over multiple frames and uses other capabilities exclusive to Nvidia's cards to stitch together the frames.

This is a different challenge than generating the content from scratch

I don't think this is possible in real-time yet, but someone put a filter trained on the German country side to produce photorealistic Grand Theft Auto driving gameplay:

https://www.youtube.com/watch?v=P1IcaBn3ej0

Notice the mountains in the background go from Southern California brown to lush green

https://www.rockpapershotgun.com/amd-fsr-20-is-a-more-demand....

All the activity around V8 Isolates + "workers" is so cool-- exceptional performance, don't need to think about regions, and CDN-like benefits

Shopify's new store-builder thinks this way too: https://github.com/Shopify/hydrogen

I love v8 isolates so far-- I'm building chat tooling with it

When added to the "edge", it means they're (insanely) fast, obliterate cold-start problem (which is killer in chat where you might have not have retry), and as long as what you write can execute between 10-50ms (with ~30s for follow-on queries) it sometimes feels like cheating

The same way Cloudflare "pushes" configuration to their network, they use a similar mechanism to push code to their edge nodes.

They have killer dev tooling too-- https://github.com/cloudflare/wrangler2

You *DON'T* need to think about regions ever-- just deploy to a lot of small regions instantly & it usually "just works" and is fast everywhere.

For extra credit, you also get access to rough-grained location information from each "node" in their network that your users connect to (globally you can get access to rough-grained local timezone, country, city, zipcode, etc): https://blog.cloudflare.com/location-based-personalization-u...

ex. for chat, could so something like to prompt for location info: https://i.imgur.com/0qTt1Qd.gif

Kenton Varda (https://twitter.com/KentonVarda) who was in charge of Protobuf and other projects gave an overview tech talk @ 10:23 speaks to isolates: https://youtu.be/HK04UxENH10?t=625

## Downsides encountered so far

- Not 1-1 replacement, think of your code like a highly-performant service worker (usual suspects: https://developer.mozilla.org/en-US/docs/Web/API/Service_Wor...)

- Many libraries (like Axios for instance) won't work since they call out to Nodejs (this might be a good thing, there are so many web APIs available I was able to write a zero-dependency lib pretty easily) They're adding bespoke support for packages: https://blog.cloudflare.com/node-js-support-cloudflare-worke...

- There's only a tiny of bit of customization for Workers required, however, there's a bit of platform risk

If you haven't tried before, definitely worthy of further examination

Re: security, it seems like a pretty good model.