laborcontract

Generate multi-shot prompts of previous requests and successful outputs.

LLMs below Opus produce a surprisingly high rate of malformed bash commands that have to be corrected iteratively or just shown the right way from the beginning using aforementioned multishot prompting.

For claude web app the best way to have the agent remember things is to have it open up a running todo list in an artifact that you ask for it to frequently update.

very poignant story about the inescapability of smart tvs

It sounds ludicrous to say out loud, but if you're staying in a Hyatt hotel, it's best not to take a hot shower until this issue is resolved. The steam from the showers tend to make these types of particle sensors go wild.

Graduating in a recession is really unfair to the graduate. Economists have looked at the effects on lifetime wages and it's estimated that recession kids lose in aggregate about 20% in lost lifetime wages, and to no fault of their own.

Of course this is an aggregate figure, but it goes to show how uneven economic outcomes can be across cohorts.

It's odd that it got second chanced pooled. It was on the front page a couple days ago. Also, the site didn't work a back then and it doesn't work now.

Flash is very very good (better than 4.1 at faster speeds). Just not anywhere near 2.5 pro.

With all due respect to the people who have studied this, I do not care. I'm relatively young, and yet the pain of typing has started to become unmistakeable and chipped away at my productivity, motivation, and sense of invincibility.

You could tell me AI coding makes me 50% slower. I'm taking it. I refuse to grind my wrists to dust.

One open secret is that batch mode generations often take much less than 24 hours. I've done a lot of generations where I get my results within 5ish minutes.

There is not a single thing in this video that addresses shoes. I want my time back.

video tldr: 3d x-rays have made bag scanners more effective at screening

Good to know. Thanks a lot!

I care much less about the training, much more about the data retention. I don't think it's wrong to not want my data retained, especially if the counterparty is receiving remuneration for the service. For free services, I agree with you. I've used the free Gemini liberally.

I do appreciate the transparency on their privacy page and their providing the ability to opt about. Seems like they've given it some thought.

Really like this. I've been trying microsoft's copilot and it's so clunky, particularly when applying edits. One would assume they have the resources to train the model..

Request: please provide a system prompt in the docs to help the llm generate the diff format that performs best w/ your models. LLMs frequently change the way they present diffs on upgrades and I don't want to be guessing which format is best.

EDIT: Please clarify your privacy policy. If my interpretation is correct, paying users will have their data retained and trained on? Is there any way to pay to use the service (w/o picking up the phone) and not have my data trained on?

  4.1 Use of Service Data

  Depending on your subscription tier:

  Free Tier: We may use your submitted code data to train our models, improve our Services, and develop new features.
  Engineer Tier: We may use your submitted code data to train our models, improve our Services, and develop new features, subject to the confidentiality provisions in your service agreement.
  Enterprise Tier: We do not use your submitted code data for any purpose other than processing your immediate request. Your code data is never used for model training or service improvement.

[0] https://morphllm.com/privacy

I use uv in my shebang and “uv run script.py”. Never run into any issues.

Here’s a couple links to discussions about it on HN:

https://news.ycombinator.com/item?id=42198256

https://news.ycombinator.com/item?id=42855258

I’ve been using and advocating uv ever since forever. It’s impossible to think about using python without nowadays.

Among many things it’s improved, scripting with python finally just works without the pain of some odd env issue.

I’ve only had boring towels in the entirety of my life and i’m suddenly curious about what your towels look like.

I wonder if this is what led to Sony forcing ATRAC down users throats. It killed their chance at any success in the mp3 player market which should have parlayed into greater success in the mobile phone market.

Crap like this can permanently alter the trajectory of a company and its products. My speculation about Google’s slowness to productize LLMs were, in part, due to the chilling effect of the Google Books lawsuit.

My first thought was wow - 20% of email startups succeed? That’s actually pretty good.

+1 to this. I was in one situation for too long. The problem for anybody in this situation is that finding a new job is hard. If you're not the type of person that interviews well or is great on paper, then the trick to pulling off the switcheroo is to focus on making great relationships now as much as anything. There are many ways to hop on to a new rail, you just have to know that it's the right decision.

Don't be a hero in the organization. In the past job. I essentially took over my manager by consistently outshining them. Then they made some unforced errors and I took advantage of the moment, made a political move, and forced their displacement. It was a distinctly unpleasant experience and something I took no glee from doing. Moreover, if you look at the grand scheme of things, nobody remembers that, and you're neither a hero to yourself nor those around you for it. It was a year or two I could have spent finding a productive, healthy, and more facilitating organization.

  100K+ Prompts Generated
  5K+ Developers


  "PromptForge reduced our prompt development time by 80%. The quality is consistently amazing."
  Alex Chen
  Senior Developer at TechCorp 
  80% faster development

There's nothing wrong with AI coding, but I'm getting really tired of seeing people post sites they've made with no attempt to screen out the most obvious of fabrications. At least it seems like you're willing to stand by it with your name.

The new UI is gorgeous and importantly, delightful.

How do you feel about him now becoming an exec?

Yes, $14B can be considered negligible, but it's a signal. My bet is that it's not a great one. Wang has been great at courting clients, time will tell us about how that translate to leadership.

One can only guess but if they were so subscribed with their existing clients why have they started focusing on much on government contracts?

Tell people you love them. It's very hard for some people to do but those three words are pound-for-pound the most asymmetrically positive words you can tell another.

There was quite loud reporting earlier this (or last?) year that Scale's data was so poor that it led to a great deal of setbacks, scrambling, and friction between the two companies. That reporting was at the top of my mind when I read this news, so I was even more baffled.

Imagine being the people at Meta who've had to deal with Scale now seeing Mark buy Scale's CEO for $14bln.

I've been all for it, look at my comment elsewhere in this thread. For me this is a huge mistake of a transaction that's neither accretive financially nor talent-wise.

This isn't the Instagram or Whatsapp transaction. Scale's been exclusively in the data labeling space.

Let's put this into perspective. OpenAI bought Jony Ive for about $5bln. Meta spent 3x that on Wang.

He is leaving to now work at Meta.

The risk here is that you've now put a person with no track record of success outside of being good at sales in charge of your AI efforts. You're betting that he's going to be the one to attract talent.

The right leadership might be able to get talent to work for a discount. The wrong one would lead to talent not coming at all.

This is by all indications the world's most expensive acquihire of a single person. Reporting has been that Zuckerberg has seen Wang as a confidant of sorts, and has proposed a vision of AI that's said to be non consensus.

Until now I've actually been a believer in the amount of money that Zuck has poured into metaverse investments. I'm not a believer of the metaverse per se, but a believer that innovation takes unafraid capex. The last thing you want to be is scared money like microsoft who chose to scuttle the hololense project over the thought of spending a couple extra billion dollars on it.

But this deal really has left me with my head scratching. Scale is, to put it charitably, a glorified wrapper over workers in the Philippines. What meta gets in this deal is, in effect, is Alexander Wang. This is the same Wang who has said enough in public for me to think, "huh?" Said a lot of revealing stuff like at Davos (dont have the pull quotes off the top of my head) that made me realize he's just kind of a faker. A very good salesman who ultimately gets his facts off the same twitter feed we all do.

On top of what makes this baffling is that Meta has very publicly faced numerous issues and setbacks due to very poor data from Scale that caused public fires in both companies. So you're bringing in a guy whose company has caused grief for your researchers, is not research nor product oriented, and expect to galvanize talent from both the inside and outside to move towards GAI? What is Mark thinking?

Zuckerberg seems to have had all the pieces to make this work but I'm a lot less confident if I'm a shareholder now than a week ago. This is a huge miss.