Hacker News new | past | comments | ask | show | jobs | submit | Oras's comments login

Quick and easy: Gemini Flash 2

More of a system: AWS Textract or Azure Document Intelligence. This option requires some coding and the cost is higher than using a vision model.


This is awesome, I was thinking of creating an agent to generate spec and sequence diagrams (using Mermaid), but this takes it to a new level being generic. I tried a question about how does LLM post training work and it made a nice workflow about it.

While it looks nice and I like the niche, I don't see how this product will survive when the same thing is doable with Cursor and Github Copilot.

Are you planning to offer your own backend, auth, notifications .. etc? Using Supabase will not keep users on your platform.


> Finding cash within the continent could be an unwitting boon if President Donald Trump’s tariff threats against European allies come to a head.

But somehow, it's easy to find cash to support wars.


*to help an ally survive against a common enemy.

Does it matter who has access to the data? It’s the principal not the actor

If there were a way to magically ensure that only the good guys had access to the information, I'd be way less concerned about these measures. (There are only a few things that I care about being secret for the sake of secrecy, and I can easily keep those off of computers.)

Every encryption backdoor is a huge vulnerability. Even if we somehow ensure that the powers-that-be remain entirely trustworthy (something that, historically, we can't even manage for a century), they're not the only people who'll have access to the backdoor. It's not possible to make an encryption backdoor that only authorised parties can use: as they say, the laws of mathematics do not respect the laws of Australia.


The actor informs the principal

Yes, from a Sunday! Is that common in legislations? Effective from a weekend?


Sounds like a being a bit in a hurry.


200k context window

$1.1/m for input

$4.4/m for output

I assume thinking medium and hard would consume more tokens.

I feel the timing is bad for this release especially when deepseek R1 is still peaking. People will compare and might get disappointed with this model.


I compared free o3 mini vs Deepseek R1 (on their website) and in my tests o3 performed better every time (did some coding tests)


The model looks quite a bit better in the benchmarks so unless they overfit the model on them it would probably perform better than deepseek.


My vibe question checking suggests otherwise. Even o3-mini-high is not as good as r1, even though it's faster than r1. Considering o3-mini is more expensive per token. It's not clear o3-mini-high is cheaper than r1 either even r1 probably consumes more token per answer.


well in my anecdotal tests, o3 mini (free) performed better than r1


Also in my coding testing o3 mini (free) is better than r1.


I did math tests. Probably you did coding.


I agree with most points, however going viral has the benefit of strong backlinks for SEO.

I had a website I shared here on HN, which made it to the front page and resulted in being mentioned in multiple newsletters and websites. The website site was on Google's 1st page when you search for the term, all due to the strong backlinks I gained.


TikTok is a product of ByteDance, I think you're mixing both.


This looks really nice and I think it’s pretty reasonable price.

I will give it a try today


Join us for AI Startup School this June 16-17 in San Francisco!

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: