definitely Anthropic would have better shot than spacex
Yes; anybody backing the “mental age of 5 guy” needs to be pretty bold, or stupid 
I would be amazed if this was not another bubble like .com, the tulips, etc. Countless billions went into central heating in the 1970s, but could you have made money as an investment? Absolutely not. AI will become another online tool, with lots of competition and everybody going crazy telling everybody how they use it to serve customers, with the corporate hype going into outer space as I write this. The reality is that the cretins at my local council are taking a month to reply to an email asking when will I get a new wheelie bin for garden waste, after they mysteriously cancelled the account and removed the previous one. Nobody will have the balls to replace the millions of mostly useless lazy people working in all the big organisations.
Tools like Claude are great for boosting the productivity of smart people.
I am wondering if I could point a camera at an LCD and have it uploading pics to Claude, so it can do code by itself It would get through fantastic money, because uploading jpegs really hammers the £££ usage. The code would also be crap, unless it is something very simple (it has trouble reading images).
For the degree of precision you need here, you can replace all trig etc functions using cubic splines
One can mostly precompute stuff, and sure there are other ways.
There’s an Info-Pilote AI available in French and English.
It only has access to the safety and medical articles published in the magazine, so the source materials should be accurate. I tried ‘ditching’ and it gave a good overview. ‘How to lean’ unfortunately recommends 50F ROP.
A gimmick in my opinion, but it might be useful for e.g. students who want to check something one level above the theory book without having to sort through all the opinions and misinformation on google.
Does anyone have any questions to run though it?
Edit: article https://www.info-pilote.fr/testez-notre-assistant-virtuel/ and link https://info-pilote.ela.amenai.net/
This is a funny sign of the times.
Anthropic has suspended its powerful new AI model after US authorities raised security concerns just days following its public release.
In a statement published on its website, external, Anthropic said it was ordered to suspend foreign nationals from using Claude Fable 5, a program that the company self-described as “too powerful”.
“The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance,” the company wrote.
Anthropic and the Trump administration are involved in a separate ongoing lawsuit over an order to stop government agencies using the company’s AI tools. The BBC has approached the US Department of Commerce for comment.
Claude Fable 5 is a version of Anthropic’s Claude Mythos, an AI program rivalling competitors OpenAI’s ChatGPT and Google’s Gemini.
Anthropic said US national security authorities had not identified specific concerns.
“Our understanding is that the government believes it has become aware of a method of bypassing, or ‘jailbreaking’ Fable 5,” the company said.
I tried Fable 5 before it was disabled for all users. It is another huge leap forward. Each time a new model comes out I test it out by building the same project – a “big sky simulator”, which tries to model, in 3d dimensions but rendered as a top-down view in a browser, like a radar display, a busy GA airfield, with arrivals, departures and circuit traffic. The aircraft fly as much like real world aircraft as possible, but deliberately have no built-in “see and avoid”, other than when taxiing and on final approach (i.e. they’ll go around if the runway is occupied). The idea is to see how long it takes and at what volume of traffic before a collision occurs. Anyway the point is, with models previous to Fable it took quite a lot of discussion with the LLM for it to generate something fairly realistic – in particular lots of repeated iterations to get the turn from base to final and the final approach to be realistic, particular with wind and inexact pilot skills being simulated. The last one I did with Opus (4.7; didn’t try it with 4.8) took a few hours of iterations with me prompting it. Fable 5 built it with zero iterations, and it’s more feature complete (for example incorporating different GA types each with different parameters), wind variation at height, etc. Whilst I say zero iterations, actually it did iterate, but not through me. I gave it one prompt (which was more ambitious in terms of desired functionality than previous ones, though I gave it far less information; previously I explained how to fly a circuit, speeds/heights/etc but this time I just said “fly a GA circuit”), it generated a detailed requirements/spec document, and then I said “ok, build it”. It took about an hour, testing itself in a browser as it went and iterating over and over, fixing bugs, rendering issues, confirming the realism, etc. This tallies with the increasing industry move away from iterative prompting and towards: here’s my goal, work on it until it’s done.
What language was it generating?
I’ve used Claude 4.8 to generate the AI portion of this
This is the C source code. Very fiddly stuff to do manually. There remain significant flicker problems caused by the background draw wiping out everything, for some tens of ms. Only the ladder is drawn concurrently with the blue/brown background. Doing the same with the arc and the text would be a lot more code. So the arc and the text get drawn in four places while the background is being drawn. But you are still left with LCD flicker due to the tearing effect.
ai_pdf
For the circuit simulator it’s Javascript because it all runs in a browser.
An intreresting wrinkle just happened to me. Out of the blue I got a mail from Anthropic telling me that my account has been blocked because there are “signals indicating possible vilation of the terms of use”. I have no idea what this is about – I haven’t even used it for several days. And all I everr did was mess about with a program for controlling a model railway, and a few things to do with my work code. Nothing remotely objectionable.
I’ve appealed, we’ll see what happens.
I would be curious to see as well what it was. I would suspect/worry about my account being hijacked and some bad actor using it…
I doubt your railway code is any type of danger for society :)
johnh wrote:
signals indicating possible vilation of the terms of use
Could that be US/non-US accounts?
Claude is just amazing.
I’ve just added a wifi subsystem (using the WF200 chip) to a product I am working on, and it took half a day. Would have taken me at least 2-3 months otherwise, of which much would have been spent digging out how it is supposed to work.
I am not even using Claude Code. That needs win10+ and I am on my trusty win7-64 which does absolutely everything and runs all tools I need. There are some hacks apparently but I use Claude in a modular manner where I specify and control the interfaces.
Out of the blue I got a mail from Anthropic telling me that my account has been blocked
Lucky it wasn’t from Amazon… A great company to buy from as we all know, but a completely useless dysfynctional disorganised idiotic cretinous cowboy outfit on the “selling to them” side The only contact is a script monkey call centre in India. No email, no post is read, and the only way I managed to finally close my firm’s amazon shop was by sending a Special Delivery letter to each of the 4 directors listed at the Companies House with UK addresses (which were of course all fake; just the head office). Took a year alltogether…
