Mark Kretschmann @mark_k

SI (AI) & tech news from a software engineer's perspective. Coding, Omarchy Linux, open source & e/acc. More building, less doom. Singularity Joined September 2007

Tweets

85K
Followers

60K
Following

788
Likes

162K

Mark Kretschmann @mark_k

a hour ago

@Jenson70089918 That's really bitter. It happened to me once that I couldn't pay in the supermarket, it was very embarrassing.

0 0 1 53 0

View Details

Mark Kretschmann @mark_k

2 hours ago

Update: Revolut and many other online banks are still having problems right now, with card payments not working at all! The Google Cloud Services outage which caused this appears to be severe.

4 4 28 5K 0

View Details

Mark Kretschmann @mark_k

2 hours ago

@patrick_y26 "AI cyber attack"

0 0 0 57 0

View Details

Mark Kretschmann @mark_k

2 hours ago

They're going to blame AI for this, aren't they?

Mark Kretschmann @mark_k

2 hours ago

Update: Revolut and many other online banks are still having problems right now, with card payments not working at all! The Google Cloud Services outage which caused this appears to be severe.

4 4 28 5K 0

6 2 30 3K 1

View Details

Mark Kretschmann @mark_k

2 hours ago

@SPiklor Great! Good to know.

0 0 0 99 0

View Details

Mark Kretschmann @mark_k

2 hours ago

@10x_alpha They need backup hosting

0 0 0 34 0

View Details

Revolut is currently experiencing an outage, affecting its app and cards! 🚨 The problems may be related to the the outage of Google Cloud Services. Details are currently not known. If you use Revolut cards, be prepared for them not working!

8 1 32 6K 1

View Details

Mark Kretschmann @mark_k

3 hours ago

Cyberpunk 2077 is getting another update! It's totally amazing that CDPR cares deeply about their games and updates them many years later!

Cyberpunk 2077 @CyberpunkGame

6 hours ago

// NEW UPDATE INCOMING… [v.2.5][22.10.26]

1K 9K 100K 3.0M 9K

1 1 23 1K 1

View Details

Mark Kretschmann @mark_k

3 hours ago

"Space Bunny" finally has a name: MiniMax M3.1 Flash. And the pricing is pretty wild: $0.15 per million input tokens, $0.60 per million output tokens, and just $0.015 for cached input. @MiniMax_AI says it cuts agent-task costs by 70% compared with its predecessor, while delivering output speeds above 200 tokens per second. It supports a million-token context window, images, coding and computer use. MiniMax reports 73.7% on DeepSWE and 61.7 on OSWorld V2, while independent evaluator Artificial Analysis also rates its capabilities highly for the price. It's available now through the API and MiniMax Code, with downloadable weights promised later. A more capable Pro version is also coming.

6 0 18 3K 0

View Details

Mark Kretschmann @mark_k

4 hours ago

@GrumpyShadows Ouch, very unpleasant experience!

1 0 1 88 0

View Details

Mark Kretschmann @mark_k

4 hours ago

@jamesfromkaw Yes, true

0 0 1 127 0

View Details

Mark Kretschmann @mark_k

4 hours ago

@VraserX Yes, for now that's true. Maybe in the future this could change, or we could get hybrids.

0 0 1 122 0

View Details

Mark Kretschmann @mark_k

22 hours ago

This is INTERESTING: Mercury, a new diffusion LLM, outperforms GPT-6 Luna, while being MUCH faster! Could diffusion LLMs become a new breakthrough for Super Intelligence?

Stefano Ermon @StefanoErmon

23 hours ago

1/4 Today, we’re releasing Mercury 3. Mercury 3 is by far our most powerful model and the most intelligent diffusion LLM ever released. It outperforms GPT-6 Luna and Haiku 5.5 on quality while running at 1000+ tok/sec.

253 404 5K 400K 2K

16 22 435 23K 134

View Details

Mark Kretschmann @mark_k

4 hours ago

@najobz Thanks!

1 0 0 163 0

View Details

Mark Kretschmann @mark_k

4 hours ago

@moubuntu Terra

1 0 1 67 0

View Details

Mark Kretschmann @mark_k

7 hours ago

GPT-6.1 Terra just surfaced in Microsoft's public code! An EF Core update merged October 16 sets "gpt-6.1-terra" as the default judge across 18 evaluation configs. Seventeen previously used GPT-5.6 Terra. The same change moves the model performing the coding tasks to GPT-6.1 Sol. The interesting detail: Terra is configured to grade another AI's work. Microsoft appears to be preparing its coding evaluations around a new generation of both models, although the configs don't establish a release date or reveal how much better Terra is.

11 10 124 11K 8

View Details

Mark Kretschmann @mark_k

4 hours ago

@ameerofthings Me too. Important to know!

0 0 1 122 0

View Details

Mark Kretschmann @mark_k

4 hours ago

@ownsomeshares Wow, not good

0 0 0 192 0

View Details

Mark Kretschmann @mark_k

6 hours ago

@notjazii For subagents it makes sense.

1 0 3 288 0

View Details

Mark Kretschmann @mark_k

6 hours ago

@Compilethings Terra is a smaller model, so yeah. Probably not.

1 0 1 322 0

View Details

Mark Kretschmann @mark_k

7 hours ago

@drakenvill87297 Then it's over for them

1 0 0 42 0

View Details

Mark Kretschmann @mark_k

23 hours ago

That's right. The sooner we can get rid of toxic "artists", the better. AI is coming for ya. The good ones will learn to use AI and become so much better!