foruhar

Apple also seems to be stirring some controversy around it's operations in Israel.

https://openletter.earth/apple-cease-funding-for-illegal-set...

The recent bombing of the bridge in Iran was apparently a double-tap strike by the US.

https://en.wikipedia.org/wiki/2026_Karaj_B1_bridge_attack

That's a fairly decent defintion of vibecoding across multiple sessions.

Hyperbolically, I think it's one of humanity's greatest resources. I can find anything from precision machining, LLM internals, historical footage of WWI, music performances from pretty much any era, and on, and on. There are so many things that I didn't know there was any footage of or that I didn't a single thing about that I find there pretty much daily.

I wish the BBC would publish their whole archive through YT. The few things that they do put up are often so mind expanding whether it's Berty Russel, The Beatles, or some cracking Scottish chap going for a bike ride with a bottle of whisky.

Happy days at the ant colony.

Smalltalk too was originally a full OS running on bare metal back in the Xerox Alto days (1972-ish).

Tesla was the real power grid guy. The scope of his invention from the generators at Niagara Falls power generation to the transformers to the motors is pretty impressive. More so given that he was eventually given the patents (originally issued to Marconi) for radio transmission.

My sense is that the Gemini models are very capable but the Gemini CLI experience is subpar compared to Claude Code and Codex. I'm guess that it's the harness but since it can get confused, fall into doom loops, and generally lose the plot in a way that the model does not in Gemini Studio or the Gemini app.

I think a bunch of these harnesses are open source so it surprises me that there can be such a gulf between them.

Gemini-3.0-flash-preview came out right away with the 3.0 release and I was expecting 3.0-flash-lite before a bump on the pro model. I wonder if they have abandoned that part of the Pareto/price-performance.

I couldn't find anything more recent that this but apparently it has made the streets safer for pedestrians too. "Traffic fatalities in the Congestion Pricing zone are down 40% from last year."

This is from July 2025: https://transalt.org/press-releases/new-data-from-transporta...

This video shows the systems being built and shipped with cooling, cabling, etc.

It’s pretty mind blowing what this crisis shows from the manipulation of atoms and electrons all the way up to these clusters. Particularly mind blowing for me who has cable management issues with a ten port router.

https://youtu.be/1la6fMl7xNA?si=eWTVHeGThNgFKMVG

This interested me as a simple-ish approach to a tough-ish promblem. From the example Trip Planner file:

"""Command line interface to process a trip request.

We use Gemini flash-lite to formalize freeform trip request into the dates and destination. Then we use a second model to compose the trip itinerary.

* This simple example shows how we can reduce perceived latency by running a fast model to validate and acknowledge user request while the good but slow model is handling it.

The approach from this example also can be used as a defense mechanism against prompt injections. The first model without tool access formalizes the request into the TripRequest dataclass. The attack surface is significantly reduced by the narrowness of the output format and lack of tools. Then a second model is run on this cleanup up input. *

Before running this script, ensure the `GOOGLE_API_KEY` environment variable is set to the api-key you obtained from Google AI Studio.

Usage: python3 trip_request_cli.py """

Nifty! Is there a github or source for this?

Iran (and some neighbors) starts the new year on the Spring Equinox, the first day of spring. It’s named Now Ruz which translates to new day. Kinda makes sense to kick off the year at spring. It’s also pretty precise give that it’s an astronomical event. It dates back to at least Zoroastrian times (15th century BCE).

All the equinoxes and solstices are celebrated there. The winter solstice is named Yalda Night, which was a few nights ago and Christmas may be related to this astronomical event. There is also Mehran and Tirgan. Ancients did like to get together and party.

https://en.m.wikipedia.org/wiki/Nowruz

https://en.m.wikipedia.org/wiki/Yald%C4%81_Night

It’s a bridge to nowhere.

The future of training seems to, at least partly, be in synthetic data. I can imagine systems where a “data synthesizer” LLM is trained on open data and probably some licensed data. The synthesizer then generates data “to spec” to train larger models. MOE type models will likely have different approaches in so far as something like a Mathematical expert likely gets a long way with training data from out of copyright works by Newton, Euler, et al.

Please say more. What was this expected and how does it support the carbon-based hypothesis?

From afar it seems like the issues around Maven caused Google to pump the brakes on AI at just the wrong moment with respect to ChatGPT and bringing AI to market. I’m guessing all of the tech giants, and OpenAI, are working with various defense departments yet they haven’t had a Maven moment. Or maybe they have and it wasn’t in the middle of the race for all the marbles.

Llaminate would be decent name for something like. Or the verb for the general wrapping of a llama compatible model into a ready to use blob.

One of the best simplified explanations of tranformers/llms that I've seen.

As a side note, drinking Coca Cola on the Tour is almost a tradition at this point.

Global entry is for gate to taxi vs taxi to gate which is TSA Pre. I find taxi to gate to be a lot slower than gate to taxi.

Very cool. When are you expecting an iOS release? Any specific issues that out iOS behind Android?

The Lathe of Heaven is worth a watch. Produced in ‘79 for public TV. Released in 1980 so maybe just outside the criterion.

Looks very cool. Is there something like this for MacOS via brew of some such vs npx?

I had access to one of the development boards 1986/87. I was writing neural network code for my degree/dissertation and it was slow on the Sun 3 at the lab which was slower than the departmental Vax. I was writing in C (rather than the AI languages of the time) to get the most out of the machines we had and it was still overnight runs to train a network. The transputer was so seductive with its promise of parallel speed but I could not get my head around the Occam language and dev environment in time for my deadline. I still think of it fondly as a thing that I was happy to have bumped into however slightly since it went the way of so many ideas about how to build a computer with an eye on the pre-winter, mostly symbolic AI (Lisp machines, Connection Machine, Cray).

Beautiful work. Even more impressive is the fact that it looks like you didn't spend much time in the UK and you have some classic British boxes. Glad to see you did the "calculator" keyboard on the PET rather than the later more conventional keyboard. Would like suggest the ZX-80 and the Research Machines 380X. Some of the early Sun, Interlisp, Xerox, and SGI machines would be fun if you wanted to venture in an adjacent direction, as would a scale PDP-11. Very well done!

Probably the weight of the cassette if we are talking about the beeb. I just recalled the “DIN to DIN” cable for the first time in 30+ years.

I'm not sure exactly what you are asking but it may be helpful to know that Switzerland is a direct democracy.

https://www.eda.admin.ch/aboutswitzerland/en/home/politik-ge....

Not only is it not getting simpler the tax code seems to keep growing as more exceptions and exemptions get added. I guess it's like so many legacy code bases that just keep growing with lots of patches for special cases.

According to the Public Law 117-154(06/23/2022), the U.S. Tax Code is 6,871 pages. When you include the federal tax regulations and the official tax guidance, the number of pages raises to approximately 75,000. This zip file (pdf) is 181MB [0].

[0] https://uscode.house.gov/download/download.shtml