m3kw9

15 ms·

421.
by m3kw9 1y ago
The ability to pretty accurately keep the same image from an input is a clear sign of it's improved abilities.
422.
by m3kw9 1y ago
To be fair, the model's ability came from us generating the training data.
423.
by m3kw9 1y ago
Type 2, they need to solve type 2.
424.
by m3kw9 1y ago
Firewall for MCPs?
425.
by m3kw9 1y ago
any one of these MCP's can have some supply chain risk where all it takes is one prompt injection to extract your chat history.
426.
by m3kw9 1y ago
You'd be surpised what people paste into the chat to ask questions.
427.
by m3kw9 1y ago
It has a check mark saying "do you really understand?" Most people would think they do.
428.
by m3kw9 1y ago
With LLMs you need to read well, they can introduce “later bugs”
429.
by m3kw9 1y ago
Is calculator/excel a bad thing? I'm ok with not having fast calculation(cognitive decline in that area) as it frees me to do other things that crops up as a result of the speed.
430.
by m3kw9 1y ago
For those not critical of what AI says, this will be a bigger issue, they will just by pass their own decision making and paste AI response, likely atrophying that thought process
431.
by m3kw9 1y ago
Aka FPGA
432.
by m3kw9 1y ago
I’m tired boss is the only response, I just stick to OpenAI or one provider as they leap frog each other every other Sunday anyways
433.
by m3kw9 1y ago
paypal is trash, if you have issues, they can hold your payments or does whatever they want to you and you have very little recourse. They make up all the rules as "regulations". A lot of times there is no choice but to use them
434.
by m3kw9 1y ago
If you read the article they actually refute it as extraterrestrial.
435.
by m3kw9 1y ago
It’s a red flag when it knows English and sent an English slang.
436.
by m3kw9 1y ago
Trust me bro theory
437.
by m3kw9 1y ago
If you are willing to pay once, you will be re-targeted. Just like Facebook ads
438.
by m3kw9 1y ago
Once you start using an npm package you are likely screwed
439.
by m3kw9 1y ago
Junior is just the lower rank, you will still have a lower rank, fine, do t call it junior
440.
by m3kw9 1y ago
Not only a prompt wizard, you need to know what prompts are bad or good and also use bad/lazy prompts to your advantage
441.
by m3kw9 1y ago
Is hard to measure return when you prompt a question and it comes out with a partial solution.
442.
by m3kw9 1y ago
is digg.com is hackernews but for everything?
443.
by m3kw9 1y ago
Ok how do I access them, or is that not the point?
444.
by m3kw9 1y ago
Money maximizers always try to scale, so we can handle a million users. I do think that’s risky most times. When you have limited resources Don’t scale till it gets close to breaking.
445.
by m3kw9 1y ago
what do you mean? For CLI or web codex?
446.
by m3kw9 1y ago
Just say begin task 1, 2 etc scroll back and see the task. Or copy paste into notes and do them sequenced
447.
by m3kw9 1y ago
Free users will get the gpt5 nano.
448.
by m3kw9 1y ago
I’m very seen ppl minds blown on free tier previous to 5. It’s basically 4o which is pretty good for normies
449.
by m3kw9 1y ago
You can use o3 for coding on plus plan almost unlimited or till they throttle
450.
by m3kw9 1y ago
Gpt5 high reasoning is a big step up from o3

More ›