Yad0

Daily Routine:
me: Wake up, pick an AI harness of the day *Clears throat*... COMPUTER! take me back to simpler times *Sobs*

[Yad0 @Micr0be]

@onusoz Sir, how exactly are you able to tell it's a "real face"? Asking for my anime gf account

[Yad0 @Micr0be]

@gregisenberg Wtf this is not true at all. Your mental model of what's valuable has collapsed into a narrow vertical slice.

[Yad0 @Micr0be]

@MattTrifilo @MrAhmadAwais Its Ad bait. Even with thousands on hardware you'll still end up with pretty much single digit tokens with no chance for multi agent setups.

[Yad0 @Micr0be]

Anyone else thinks grok 5 is going to be a big leagues disrupter?

[Yad0 @Micr0be]

@CtrlAltDwayne This 100% this and engagement/rage bait, I hate them and aggressively unfollow (that's how I engage)

[Yad0 @Micr0be]

Unless of course you tell the AI to "make no mistake"

[Scott Stevenson @scottastevenson]

Make no mistake: the state of AI writing is really fucking bad

[Yad0 @Micr0be]

@cherry_mx_reds No it's not, stop feeding local AI hype so poor basterds spend a shit ton of money for single digit tokens per-second

[Yad0 @Micr0be]

@pmddomingos What a pile of bullshit. If you know anything about anything cyber then you know this isn't true. You're either an anthropic hater or a retard, maybe equally both.

[Yad0 @Micr0be]

@fchollet More ppl know this than twitter makes it seem, it's either those who know don't bother vocalizing it or those posts are buried under a megaton of engagement bait and slop posts.

[Yad0 @Micr0be]

@0xSero Good model, but for local, just no.

[Datacurve @datacurve]

GLM 5.2 is now on DeepSWE as the top open-source model on our leaderboard.

With a pass@1 score of 44% at max effort, GLM 5.2 is indisputable #1 open-source model besting Kimi K2.7 Code by 17%.

[Yad0 @Micr0be]

What you see when you step outside is not governed by technological advancement or futuristic change, it's governed by bureaucracy and that is stuck in the 1980s era and can even be argued it has regressed.

Wherever bureaucrats have conceded you can immediately see the difference, which is rare.

[Yad0 @Micr0be]

Google is behind even the Chinese labs in the AI Race (trust me I know how far behind they are), but if you are a gambler you would bet on them for the long run

[Yad0 @Micr0be]

Easy, because they targeted different users and Anthropic built more focus and gathered momentum around Business/Enterprise and that was mass CEOs PR driving force that fuelled the mainstream news/hype cycles.
Regular users and high quality user experience OpenAI was going after were too sluggish and far too small of a productive user base to feed the media in the same way.

Even now CODEX is behind Claude Code in Enterprise domains, and it sucks because IT IS better.

I spend my days reminding one of the biggest orgs in the world that CODEX is better and GPT is cheaper.

[Yad0 @Micr0be]

@nutlope can you share the type of prompt you use to get this style of dashboards please?

[Yad0 @Micr0be]

@icanvardar This was predictable over a year ago if you keep an eye on Compute. (They sold most of theirs a long time ago, that's why they are crawling it back by buying from SpaceX)

[Yad0 @Micr0be]

my GPU has 16GB VRAM and CPU 32GB RAM. damn it!so close! What's worse is every X account promoting local models seem to have a data centre or 2 under their beds 🤔

[Zixuan Li @ZixuanLi_]

GLM-5.2 can now run locally in llama.cpp and Unsloth Studio. Check the graph below for the accuracy of each GLM-5.2-GGUF quantization.

Full guide: https://t.co/ZcSbQk7Aqv

[Yad0 @Micr0be]

@briandecoded @petergyang haha same🤣

[Yad0 @Micr0be]

@petergyang @briandecoded he probably said "make no mistakes" so it one shots everything

[Yad0 @Micr0be]

@Ceoz_1 @0xSero all of them, at some point they just ramble go off the rails. That's why that bar graph looks the way it does.