heymax054

Semi-related: I posted this yesterday: https://news.ycombinator.com/item?id=45828917

Different models perform differently when it comes to catching/fixing security vulnerabilities.

Of course. Are they useful enough though for people to install an entirely new software?

90% of their features could fit inside a VS Code extension.

There are already a few popular open-source extension doing 90%+ of what Cursor is doing - Cline, Roo Code (a fork of Cline), Kilo Code (a fork of Roo Code and something I help maintain).

No need to change your IDE to AI/vibe code.

There are already a few good VS Code extensions like Cline and Kilo Code which do 80/20 of the job.

I just use Anki (spaced repetition) to learn new concepts. Now with AI, one added benefit is “avoiding skill/knowledge atrophy” the more I use LLMs to generate the clde.

$15 worth of Claude 3.7 tokens? Does this translate to 3 “hello world” files?

There’s been a post on the HN frontpage about “vibe coding” almost every day, so pretty sure many people are familiar with this term. But you have a point, will add this link to the post description.

Edit: Updated the question with this link for those who don’t know what “vibe coding” is.

Interesting. In the future we might even see influencer "template marketplaces" where you can buy the rights to use an influencer's face (and pay additional if you want them to promote that image on their profile).

This will probably change the default perception. Nowadays, if people see something online, they assume it's real. The reverse will prob be true in 10 years.

Again, there ARE definition changes.

From [1] (an article you referenced):

> When a person has more than one account and takes > actions (such as liking photos or adding comments) on the separate accounts, these actions may be counted separately even though they were made by the same person.

Notice the "may be counted" part (which is pretty vague and it's hard to argue the definition hasn't changed on a vague statement). Judging by the their latest announcement [2], by mentioning "may be counted" in [1], they probably meant this:

- If that persons' email was the same on two or more accounts, or they used the same mobile device, we'll count it as the one person

- Otherwise, we'll count it as a different person

Again, I'm quoting this part from [2]:

> For example, if someone used the same email address across their Facebook and Instagram accounts or accessed both platforms from the same device, we counted them as one person when they interacted with ads.

The definition of "may be counted" will essentially change after Facebook implements their counting update to:

- If a person didn't link their accounts under "Accounts Center", we'll count them as 2 people

- Otherwise we'll count them as one person

So Facebook won't make any attempts to implicitly associate 2 accounts as 1 person using emails/device IDs as they did in the past.

Based on the available data, the definition of what Facebook counts as a "unique person" HAS changed.

[1] https://www.facebook.com/business/help/283579896000936

[2] https://www.facebook.com/business/news/an-update-on-how-we-c...

> Advertisers can clearly see how this is calculated.

Do you have any evidence that the average advertiser can "clearly see" how this is calculated? Try this: Go to any Facebook Ads group, DM 10 advertisers and ask them if the "Impression" metric tracks accounts or people. See what % of them reply confused.

Second, it appears that you haven't ran Facebook Ads at all. When you see your ad set performance, one of the primary columns you see is "Reach". If you hover over, you'll see this:

> Reach > The number of people who saw your ads at least once. Reach is different to impressions, which may include multiple views of your ads by the same people.

My argument is that Facebook is changing the definition of "people" to "accounts". If you are the SAME ad both on Facebook and Instagram, Facebook will report your reach as "2" (compared to "1", as it was before the change). How is this "noise about nothing"?

Third, the documentation you linked me to has been recently updated to reflect the news release Facebook announced. This information hasn't always been there.

Your arguments made sense at the beginning (with regards to the "double pricing" argument) but they appear extremely biased now.

Removed the price reference to focus on the article's main point. Honestly it's a bit frustrating when you've spent hours researching something and reached the HN front page to have someone tear you apart over a small speculation at the bottom of the piece. But that's HN, I guess.

Thank you for the correction.

Here: https://www.facebook.com/business/help/1461718327429941?help...

Thank you for pointing this out, though. The purpose of the article was to make people more aware of the change and how it can potentially mislead advertisers. I edited the section you quoted to make things more clear.

It can be tied to humans/accounts. Facebook has a "cost per 1000 people reached" derived metric: https://www.facebook.com/business/help/1461718327429941?help...

Facebook changing their definition of what a person is (outside of the normal societal definition) was the real issue here.

EDIT: Removed the "double" and the whole price reference because it caused confusion.

That was the top comment on this HN thread:

https://news.ycombinator.com/item?id=28702997

The reply was: "Per operation charges, even if at the same level as EC2, would still be substantially cheaper."

I've been wondering on the pros and cons of interviewing customers vs. just putting something out there and seeing the reaction/getting a commitment.

There's a great book about the latter called "The Right It" by Alberto Savoia (a guy who used to work at Google to test business ideas). There he describes 8 different ways to test a biz idea:

- The Mechanical Turk – Replace complex and expensive computers or machines with human beings.

- The Pinocchio – Build a non-functional, “lifeless”, version of the product.

- The Minimum Viable Product (or Stripped Tease) – Create a functional version of it, but stripped down to its most basic functionality.

- The Provincial – Before launching world-wide, run a test on a very small sample.

- The Fake Door – Create a fake “entry” for a product that doesn’t yet exist in any form

- The Pretend-to-Own – Before investing in buying whatever you need for your it, rent or borrow it first.

- The Re-label – Put a different label on an existing product that looks like the product you want to create.

A ques to the author (I saw he commented here). What do you think are the pros and cons of interviewing people using the MOM framework vs. just putting something out there and getting feedback? Often people give very different feedback when you ask them questions vs. present something and ask for skin in the game.

I wonder if this will impact productivity in an unexpected way.

When you disclose the numbers by which you "track" people, people will find ways to maximize those numbers without necessarily maximizing output.

But then, is there a good alternative? Big tech asked for this by being extremely non-transparent and mistreating people.

While I agree with the intent, I don't think simple solutions like these are necessarily the best ones to solve the problem.