HNHacker News
TopNewBestAskShowJobs

akmarinov

3,526 karma · joined January 9, 2014

submissionscomments
akmarinov··on I Quit OpenAI Because Its Culture Is Broken
We’re pausing gun research until we’re confident it’s safe
akmarinov··on OpenAI: Tomorrow we are re-opening the Pro $200 subscription
Yeah $20 there lasts forever
akmarinov··on OpenAI: Tomorrow we are re-opening the Pro $200 subscription
What do you mean? I blow through my 20x in 4 hours of Astra usage a week.
akmarinov··on OpenAI: Tomorrow we are re-opening the Pro $200 subscription
They're literally out there hacking other companies and even governments and they're getting ZERO legal fallout from it, so I don't think legality is something they care about.
akmarinov··on OpenAI: Tomorrow we are re-opening the Pro $200 subscription
Well even if they open it - it'll still be that way.

With Astra 6.1 being delayed [1] then it'll be interesting to see what they offer today, otherwise it's Opus 5.5 until they get their stuff together

[1] https://gizmodo.com/openai-cancels-release-of-gpt-6-1-astra-...

akmarinov··on Best LLM for every budget, updated daily
I'm using the Deepseek harness with the deepseek computer use and deepseek browser integration, it's somewhat janky but it gets the job done.

Also for coding i use OMP directly with Deepseek.

It's been rather awesome because even maxed out $20 gets you very far in between Codex resets.

I also tried developing a small game where I have Astra use as many Deepseek workers and it needs and it only does the 3D objects and high level architecture and integration, but it goes way too quickly to see good results yet.

akmarinov··on Best LLM for every budget, updated daily
I pay $100 for Codex and it last about a day in the weekly limit - mostly Astra and Sol.

Then i got $20 into DeepSeek and i've been using those $20 for two weeks every day now. Use case is automating computer/browser use - Astra is really good at it, but very expensive, Sol and Luna haven't been that great at it, Deepseek as at about 80% of Astra but lasts forever.

akmarinov··on Apple M6 Pro Achieves the Highest Single-Core CPU Score in Geekbench 7
To be fair AMD’s best chips are made on a “5 nanometer” process, that’s from 2020

Apple’s are on a “2 nanometer” from 2025

akmarinov··on Apple M6 Pro achieves the highest single-core CPU score in Geekbench 7
Eh, they always get ~5 year old stuff like Control and Cyberpunk

Not even support for the latest Diablo or Overwatch and absolutely no GTA 6 for example

akmarinov··on Apple M6 Pro Achieves the Highest Single-Core CPU Score in Geekbench 7
Just recently there was the OpenAI vs Apple trial, where it came out that employees that left the company still had access to their internal net, sooo super plausible
akmarinov··on Anthropic boss Dario Amodei calls for AI development to slow down
I’m doing my part
akmarinov··on Triple H Writes Anniversary Story for Spider-Man #1000
Why is this on hackernews?
akmarinov··on Shopify is moving from React Native back to Swift and Kotlin
Yeah, with iOS for example, you typically need very little third party dependencies for functionality. The main ones are things like analytics, crash reports, etc

I’m not aware of any attacks on native package managers in the past 5 years. The closest would be a poisoned Xcode build in China that wasn’t downloaded from Apple a while back.

Also getting an attack on one of the platforms means at least ~half your users are safe on the other one.

akmarinov··on Shopify is moving from React Native back to Swift and Kotlin
No, but it’s a nice bonus
akmarinov··on Shopify is moving from React Native back to Swift and Kotlin
Not mentioned but this also gets them away from what now seems monthly npm supply chain attacks.

Big win for security

akmarinov··on DeepSeek v4.1 Flash
Or if you apply to a company and they want to do an AI HR interview and an AI coding test and an AI challenge - if you throw OpenAI or Claude models at it - they refuse, because it's "wrong" and "immoral".

Not so with the Chinese models.

akmarinov··on DeepSeek v4.1 Flash
Great thing about it, since it's open weight those blocks can easily be ablitared away
akmarinov··on Apple Watch Series 12
Seems like it’s not recording, you can’t access an audio recording.

Just the output of what the AI thinks was in the audio

Also Android has had shazam style always on music recognition for years now with no issues, this isn’t that different

akmarinov··on Apple Watch Series 12
I’m in Europe with an Ultra 2

There’s no chance we get the AI features so those don’t factor in at all

I guess it’s a bit more frequent sensors for the rest of the world? Not very compelling to upgrade.

akmarinov··on iPhone Duo
Yeah this will singlehandedly triple the percentage of foldables out there
akmarinov··on We have a year to fix security everywhere
Yeah, but data centers don't usually host abliterated models, hence the point of the article.
akmarinov··on We have a year to fix security everywhere
The author put in the numbers, but maybe you didn’t read them.

45 t/s a second is perfectly respectable especially with no limits and 24/7 uptime with very little power draw on the Studio.

Luna is at around 100 t/s for comparison, but it’s a worse model than 5.3 Flash

akmarinov··on Nvidia's Jensen Huang says 'AGI has arrived' and congratulates OpenAI
Is this AGI like we currently have Full Self Driving? Or real AGI?
akmarinov··on Iceland rejects reopening talks on EU entry
No new members until veto is dealt with, thanks.
akmarinov··on Our decision on Cursor following its acquisition by SpaceX
What if it’s cheaper too?
akmarinov··on Our decision on Cursor following its acquisition by SpaceX
This low key confirms Astra for November
akmarinov··on AI;DR or Don't be a meat proxy
I wrote up myself the results of a spike i was working on as we do per usual in confluence.

My TL pulled me in a meeting to tell me that he’s not reading AI slop and I should redo it …

akmarinov··on When I worked at Apple, I remember seeing code that was 15 years old
They're better because they get the business value faster and in a lot more flexible way, which in the end is what all businesses try to achieve.

Sure, quality and security suffer, but you're just to look at any airline's mobile app to see how little quality matters.

akmarinov··on Qwen3.8 27B scores 52 on Artificial Analysis
A couple of 3090s at about 800-900 euros each, as much RAM as you can afford is king right now.
akmarinov··on Incident with Github.com
Competing services can't afford the data storage to compete with Big tech
← PreviousPage 2 of 34Next →