TheJCDenton

Très sympa

There's something quietly poetic about a thread on declining cognitive performance where one of the top comment is "here's a video if you don't want to read."

Very informative, thank you. I was about to buy one of their products but this made me pause.

What you think would be the alternative in Europe ?

A little more context : in june 2024 at Computex, Tuxedo announced a possible christmas 2024 release [1]. A Qualcomm/Tuxedo collaboration was expected but did not materialize [2].

[1] https://www.tuxedocomputers.com/en/TUXEDO-on-ARM-is-coming.t... [2] https://www.tuxedocomputers.com/en/How-is-TUXEDOCOes-ARM-Not...

The main metric has always been care.

Care beat quality as a metric, because care is very inefficient to fake and very powerful when genuine.

How does the company providing the physical product or the service care ?

Most companies right now care about AI. Some integration are impressive. But where's real care about users ? It seems it's not the subject anymore. We may have tricked ourselves into beveling technology will resolve in itself all problems, and it's at its peak with AI. As engineers we can forget sometimes that technology is just a tool and its fine, but as a society it may leads us all in a bad direction.

Very nice, thank you !

It seems that another round is necessary, and VCs want AI. You and the CEO are competent and transparent. This is the premises.

You just need to have the conversation with the CEO, but in the right order.

Come to him/her with objective data, but without your conclusions or concerns, just your observations about where this is headed and if he sees at what rate this AI focus is eating ressources. Ask the CEO what it means for the company. Let him/her do the talking.

From this point, the CEO have the data and you have the company policy. Know you can talk on equal grounds on the matter with him or her. Good luck.

Haven’t played it yet, but if it’s good, I hope the eerie world of Morrowind is next !

What if high quality social interactions are not the norm for some kids, and these systems are giving an inspirational blueprint for what they crave and don't have access to ? In disfunctioning families, this can be a lifeline.

Last LLMs launches from top US AI labs were rather disappointing. Products seems under baked and mainly reactive... They oversold the capabilities of their future models when Deepseek landed to stay in the news at the time, but now that we look at the result it's not what everyone expected. And people are starting to question the hype, which is healthy but also very dangerous for these companies needing very large capital. It seems that some labs like Meta are unto something, but it's more research material for now than a short term product. Interesting times.

Extremely useful

Unfortunately, this might be the appropriate tone for dealing with the U.S. administration.

This Deep Whining® technique used by OpenAI is not very effective.

Very interesting take

There is an editable tasks list and in the settings menu you can choose to receive notifications via push and/or email.

Nothing yet

It seems to demonstrate that if you're frail from the start it's statistically a bad omen.

The new version finally enable variables in API by environment, dependencies between your own ff projects, and come with a VSCode extension, among other things.

Widget Builder: Build incredibly flexible components by passing components as parameters, allowing for endless customization and reusability. Dev Environments: Hook up multiple databases to the same project, or reference environmental variables in API calls or other parts of your project! More Action Triggers: Unleash a wave of new interaction possibilities with additional action triggers like keyboard shortcuts and new gestures! Better Collaboration: Work seamlessly with your team using improved branching and commit features, making version control and collaboration a breeze. Import Projects with Libraries: Effortlessly reuse UI components, data types, app states, and more by publishing and using Libraries. VSCode Extension: Enjoy a more robust code editor experience with seamless integration with Visual Studio Code.

Cursed

> We are problem solving addicts. > Without that fix, we get [...] anxious [with] just like normal withdrawal symptoms.

Absolutely.

Solve problem -> Less uncertainty -> Less anxiety

It seems to me that's exactly the positive feedback loop we are hooked on. And the more your are capable, the more you can be subject to it. Which is fine until you realize you will also suffer more from its inevitable reversal (stalled problem -> more uncertainty -> more anxiety). So you find a different but even harder problem to solve to get the high... Deep inside you're still insecure and hurt, but on the way you may build a digital payment processor, sell a lot of electric cars and start a space rocket company (but your mileage may vary).

Joke aside, there are elements of mental health that must not be overlooked when playing these games and working hard all days long ; but if it's what that make you cope with life, have fun and make a net impact on society, it's still a win.

Intellectual output is not lost, it's just maintenance for people with brains wired a certain way. Without these games the intellectual potential would surely burn/bore out the people actually wanting to play those games, as real life may not give access so easily to so many interesting problems.

AWS SAM. I'm using Go.

If you have any architectural complexity in your program, good error handling often come with custom errors. Unique error strings just don't cut it, you need a better structured type than a string to apply any kind of logic.

I have been in a similar situation. They are incredily picky, maybe because a few years ago, they were incredibly accepting. The minimum would be that they share their criterias, or at least enough details for us to infer what would be a deal breaker for them. As of now it's not sufficiently clear. Good luck.

Finding printable characters in some archives or even machine code can be done without decompiling anything. Unix tools like strings[1] are usually used for this kind of work.

[1] https://en.wikipedia.org/wiki/Strings_(Unix)

The quality is not great nor terrible, but with a new and huge code base it helps find a starting point to work on it and to understand what's going on.

The demo [3] seems very promising.

"Their method cleverly exploits the LLMs' tendency to use initial tokens as "attention sinks" to anchor the distribution of attention scores. By caching initial tokens alongside recent ones, StreamingLLM restored perplexity and achieved up to 22x faster decoding than prior techniques." [1]

"We show that StreamingLLM can enable Llama-2, MPT, Falcon, and Pythia to perform stable and efficient language modeling with up to 4 million tokens and more." [2]

"we discover that adding a placeholder token as a dedicated attention sink during pre-training can further improve streaming deployment." [2]

"StreamingLLM achieves an impressive speedup, reaching up to 22.2× per token. Despite its reduced latency, StreamingLLM sustains a memory footprint consistent with the re-computation baseline." [2]

[1] https://notes.aimodels.fyi/llm-infinite-context-window-strea...

[2] https://arxiv.org/pdf/2309.17453.pdf

[3] https://github.com/mit-han-lab/streaming-llm

1. Recording ceremonies like sprint plannings. 2.a Using whisper to go from speech to text 2.b Playing with diarization solutions to assign text to people 3. Still playing whith chatgpt to summarize all of this, make categories (people, projets, pain points and the like)

As i enter a new mission, this is very very usefull.

First, i'm sorry for your son.

I've been hacked a few years ago. But it was in the several tens of thousand of dollars (for me in a matter of hours). I've deactivated my account while the hacker was using it.

It took a stressful week to settle the matter with AWS. And not at any point i've been in a position to influence anything. I was just there for the ride. In the end they erased the debt.

A few years after that, i've reactivated my account. Some leftover resources were there, so i cleaned but i missed some, resulting in another charge, less expensive. I've requested to be reimbursed. First they objected, but then with some references to the previous incident i gave them they complied.

Do not give up. Do not pay.

IMO, more static code means even more predictability in terms of object placement in memory, thus allowing much more aggresive memory optimization : less fragmentation, efficient pagination etc... It means also that you have static code and GC interacting with it, it's pretty new, isn't it ?