kingkongjaffa

6 ms·

1.
by kingkongjaffa 3d ago
> When the purpose is conveying personal thoughts, AI should be used lightly. When the purpose is to convey information, it can be used extensively. I would argue that this is very wrong. I think it's wrong because when we write, ev
2.
by kingkongjaffa 6d ago
up until GPT 5.6 sol the answer was Claude. Since then GPT5.6 Sol and GPT6-Astra are better than Sonnet/Opus/Fable. IMO as someone using and building with these every day.
3.
by kingkongjaffa 7d ago
You say that, but I've bought the non-pro iphones the last few generations (13, 15 currently) and consistently wanted a better camera, so this upgrade cycle I'll probably shell out for the 18 pro.
4.
by kingkongjaffa 7d ago
I hope, not I will never buy airpods with cameras.
5.
Ask HN: What is your best Agent Skill?
1 points by kingkongjaffa 8d ago|2 comments
6.
by kingkongjaffa 10d ago
Absolutely idiotic take. Why is the Guardian peddling this opinion piece? Stepping down as CEO would do nothing, so unless you plan on forcing him to sell his share IDK what the point of the article is. Mark Zuckerberg (13.68% equity; 61.2%
7.
by kingkongjaffa 11d ago
I used the same prompt in 5.6 sol and in Astra Astra did a smarter web search and found higher quality resources and blog posts about the topic I was trying to learn. On the same effort level, while it was cranking through it did about 3-5x
8.
by kingkongjaffa 13d ago
For work I use a work provided laptop, mostly folded on my desk with an external monitor and keyboard and mouse. At home I use a Macbook Pro. Because of the RAM cost rises and desire to run LLMs locally, I have considered a mac mini or some
9.
by kingkongjaffa 14d ago
> this point earns its keep. This and 'load-bearing' anywhere are just awful. Anthropic would do well to stop inserting their poor writing taste into models.
10.
by kingkongjaffa 17d ago
Talking to customers is the most important thing, not sure prompting really matters at all, LLMs lack true novel creativity and cannot prioritize product features based on direct customer value.
11.
by kingkongjaffa 17d ago
> I consider my prompts and prompting style to be trade secrets. No one wrote this with a straight face did they?
12.
by kingkongjaffa 17d ago
This looks okay: https://inferenceengineering.tech/ Although I took a few of the quizzes and thought several of the questions were opinion rather than factual i.e. several possible correct answers are shown in practice. Thi
13.
by kingkongjaffa 17d ago
- turn off memory and delete - turn off chat search so it cannot search other chats for context - delete custom instructions / system prompt - do not run inside a claude project or GPT project - delete all skills and plugins Most sourc
14.
by kingkongjaffa 21d ago
Isn't there always a political angle in the act of creating software? In essence the human right and ability to learn and change and wield computers at an unlocked and low level is a policy decision. What if governments said programs m
15.
by kingkongjaffa 24d ago
> find the cheapest listing of the product your business need Think you need to narrow down what is exactly being purchased. Not sure whether you mean something like 'find me a cheap SAAS for CRM' and it comes back with quotes
16.
by kingkongjaffa 25d ago
> Changing it to "and" would make it sound more natural To whom? > Your choices are tea or coffee. is the obvious way to write this as english speaking person talking to another, 'choices are tea and coffee' is jus
17.
by kingkongjaffa 28d ago
give them the link to github? if you need authentication there's not really any way around that, you could implement SSO or something but that would require more collaboration with your engineering/IT department.
18.
by kingkongjaffa 29d ago
Is this through LLM providers on a metered API connection? You're probably at the scale that having your own hardware to run local models is worth it, even if it's a couple of mac minis or mac pros in a cupboard in your office.
19.
by kingkongjaffa 29d ago
I also cancelled Claude recently. GPT 5.6 models are really good, and they have much better usage limits.
20.
by kingkongjaffa 1mo ago
I think it's fair to say they started off bad, but redeemed themselves. In contrast Star Citizen is far more ambitious but shows 0 sign of ever actually landing. NMS is still a great fun space game to play, the lore / back story
21.
by kingkongjaffa 1mo ago
> when high quality online education was provided for free Is it not still happening?
22.
by kingkongjaffa 1mo ago
No way of knowing, this report is about revenue not profit.
23.
by kingkongjaffa 1mo ago
I love this article, it's been posted here several times before. The first time I read it I had been in New York and was waiting in JFK to fly home, so naturally it resonated. In some ways it's THE perfect hacker news article. It&
24.
by kingkongjaffa 1mo ago
I've done all of the note taking stuff, evernote, todoist, trello, emacs org-mode, notion, obsidian. Ended up on https://bulletjournal.com/blogs/faq/how-to-start-a-bullet-jo...
25.
by kingkongjaffa 1mo ago
As a day 1 claude code user I've been a Claude fan boy but Opus and Fable usage and pricing is just not viable. I cancelled my personal claude account and got chatGPT $20 - the amount of usage is such that I've never run out ever,
26.
by kingkongjaffa 1mo ago
Yes exactly like that.
27.
by kingkongjaffa 1mo ago
GPT5.6-Sol is brilliant on Max in my experience.
28.
by kingkongjaffa 1mo ago
Yeah, immediate thought was why should a 4 year old device be in need of any revival? Specifically a writing tablet, I assume it just works still. But after reading the article the scourge of 'always online' devices seems to have
29.
by kingkongjaffa 1mo ago
> Usually once MR is merged No, nothing is merged yet I'm talking about during code review when changes are requested by the reviewer, you force push to overwrite your previous commits, the brand history does not know this by design
30.
by kingkongjaffa 1mo ago
how are you measuring ROI? Are LLM products driving revenue?

More ›