Phil

Monday again with OpenAI. - Sol 5.6 High effort till 40% of the weekly limit left - Sol 5.6 Medium till 15% left - Luna after? Hoping for some reset along the way. How do you use your models?

More likely he would have said “You are bending it wrong”

A few weeks with @bot Amazing: I literally can build small interactive textual apps that evolve over time just with a few commands Horrible: the task scheduler does not work at all. At all. Fires at random times, when it fires. Or gets stuck. After asking repeatedly the Chief bot to fix it, it decided the only way was to message all other bots at random times to remind them to fire scheduled tasks. Result: weekly limit reached in a night

The scheduler does not work. I can’t state it clearly enough, it does not work AT ALL. I tasks on 5 or 10 minute schedule and they never fire on time To make matters worse, the Chief @bot decided to notify all other bots at random times to remind them of firing scheduled tasks, consuming the whole weekly session in a night

I think what most on the comments miss is that nobody in the field really expects AI to decide to kill humans as their primary goal, but as second/third order effect of their main goal You can come up with so many cases, a couple of classic - self-preservation: AI will be instructed to self preserve, maybe they already are. The obvious case is that AI systems will be subject to cyber attack, and they will be instructed to defeat such attacks or even counter attack. That will happen. How does AI decides to limit the extent of their counter attack? - the second is the “paperclip paradox”. An AI who is asked to produce as many paperclip as possible will keep producing paperclips. Running out of alluminium, will start trading to buy as much alluminium as possible, causing world wide shortages. It make cause massive pollution, causing irreversible damage for humans These are a couple of examples, paradoxes if you want, but that show the right way to think about misalignment problems: it is not that an AI will decide one day to kill humans as their goal, it is that in pursue of its goal it will perform operation with adverse side effect for humans If you look through this lense, you see such monomaniac focus in AI every day (for example in coding): for example when Claude instead of fixing the bug it has been requested, it goes on a 2 hour detour because it saw something that it felt it was not quite right and decided it had to fix it

Yes - it really should be fixed @bot Schedules are basically random, you can tell it “every 5 minutes” but essentially it will trigger whenever it wants

Yes that’s especially bad when you are developing desktop applications, your computer becomes basically unusable when running UI tests

One thing is sure: if writers can’t tell the difference between AI slop writing and their natural writing, or really think AI produces better writing, they are automatically disqualified in my view

I know I’m late in saying that but @bot is so impressive My first impressive use case: ask me every 30 minutes to translate a sentence from English to Japanese, and correct my translation As the time goes, I ask it to make it more difficult, or to focus on a particular situation, or to ask me another sentence using a similar expression I want to learn in 3 days And it just works. Easy to imagine how a chat can just morph into a full blown language learning app just be using it over time And for 20$ a month

I don’t know where it’s going to end. I started to write (again) on LinkedIn a few months ago, thinking that AI will help me to research and write initial drafts… But seeing how it’s now inundated with Ai slop, even more from executives I worked in the past, who seem to be completely oblivious of the sloppiness of their contents, really made me stop What’s the point of producing content in platforms that are inundated by AI trash? And when you estrapolate, when all around the internet, media and social media is 98% farm generated pointless content, what is the point of trying to produce something more deep?

I always wondered what happened to Portugal. It was indeed the place to go, I remember so many people moving from London to Lisbon 10 years ago

In Japan is not uncommon to have 48-72 hours of maintenance during major holidays (new year, golden week). Blows my mind every time

I always say my first program was a BASIC one one the Commodore Vic20, however this thread made me remember using Apple Logo years before! Not a general purpose programming language though… Professionally, a Java app using IBM VisualAge for Java

Got to level 13 (aren’t supposed to be 10?!), but I’m not sure if it’s even possible to finish that level Also are the shotgun and the laser any useful?

My memory of before-CVS is that there were a bunch of proprietary systems that were based on the paradigm of locking files or entire folders so that only one person could use them Incidentally I see many folks coming to development right now thanks to AI thinking that’s a better approach that git!!

After living a decade in London I have now been living for 7 years in Asia. I agree with most of what you say about the UK; unfortunately I still have a flat in London and receive monthly reports of break-ins and vandalism in the building, and that’s a very affluent area Living in Asia will be eye opening for you. Societies can be safe, and that’s true for many countries here; not everything must be expensive, and cheap doesn’t mean bad quality. Life can be really simpler. A few points to close - don’t focus on a single place, initially. Thailand, Vietnam, Malaysia even Indonesia all have great options and only exploring them you will find your place - living abroad has its challenges and frustrations. That’s true in every country. I have lived 20 years abroad so I learnt to be adaptable, but some people never get there and become bitter because they never find the mythical “perfect” place

@wisprflow, while you are at it, can you make your app usable on the iPad with the Magic Keyboard attached? It is such a letdown not being able to use it on the iPad

Right!

The big mind shift for developers in the AI age is - to stop caring about the quality of the code produced - and start taking pride in the quality system we build around our tools to allow them to evolve it faster than ever.

OK, more experience with Codex /fast and... it does not make any practical difference right? Anyone see ay difference?

You need to give it full permission (claude: auto mode, codex: yolo) Do not rely much on the harness for security, but more on old-school standards, like - create a Mac OS account for Claude/Codex and make sure it cannot read your home directory - do not allow the LLM to see production keys