Falcon F-14 🇺🇸

If you can keep it...

[Falcon F-14 🇺🇸 @BlueMaxSystems]

Terrific resource!

[Ahmad @TheAhmadOsman]

This should be read in conjunction with the article below & you'll be able to see the full picture of local AI

https://t.co/KCSgrCWeyd

[Falcon F-14 🇺🇸 @BlueMaxSystems]

Friday night and right on track @claudeopus is nerfed for the weekend. Time to focus on the local models until Monday!

[Falcon F-14 🇺🇸 @BlueMaxSystems]

The latest updates from @anthropic @Claudeopus_AI ...

"⏺ API Error: Server is temporarily limiting requests (not your usage limit) · Rate limited"

[Falcon F-14 🇺🇸 @BlueMaxSystems]

I have gotten this 4-5 times last 2 days. Anytime harness or agent memory is the topic... "Fable 5's safety measures flagged this message for cybersecurity or biology topics. They may flag safe, normal content as well. These measures let us bring you Mythos-level capability in other areas sooner, and we're working to refine them. Switched to Opus 4.8. Send feedback with /feedback or learn more: https://t.co/Pyc5JTSpli"

@AnthropicAI
@fable

@claudeai

[Perry E. Metzger @perrymetzger]

I have seen some people claim, completely incorrectly, that Anthropic’s new “silent sabotage�� policy is just trying to stop distillation attacks that steal weights from their models. This paragraph makes it extremely clear that that is not the case. Anthropic is generally blocking attempts to advance the state of the art in AI research, even for techniques they themselves make no use of.

For example, Anthropic is targeting research on distributed training, which is to say, training runs done on large pools of computers loosely attached over the Internet, but it does not do any distributed training, as it has no need for it. Distributed training is, however, a technique that open source AI researchers without billions of dollars at their disposal have been leaning on to allow them to leverage large amounts of crowdsourced computation.

Why would Anthropic do this? Likely because they fear that their model might help people write code to build open source AI systems trained without expensive hardware, and because to the EA cult, distributed open source AI training systems are the worst possible thing someone could build.

It is perfectly within Anthropic’s right as a company to do stuff like this, but it is also within our right, as users of AI systems, to stop giving them money.

It is perfectly within Anthropic right to lobby to try to gain a government monopoly on AI research and development, as they have for years now. It is perfectly within our rights, as citizens, to oppose repugnant attempts to grab power with all our strength.

[Falcon F-14 🇺🇸 @BlueMaxSystems]

I have gotten this 4-5 times last 2 days. Anytime harness or agent memory is the topic...

"Fable 5's safety measures flagged this message for cybersecurity or biology topics. They may flag safe, normal content as well. These measures let us bring you Mythos-level capability in other areas sooner, and we're working to refine
them. Switched to Opus 4.8. Send feedback with /feedback or learn more: https://t.co/Pyc5JTSpli"

@AnthropicAI @fable @claudeai @theo

[Theo - t3.gg @theo]

Mythos is here. It's called Fable 5. I spent the last 24 hours spending thousands of dollars in tokens to give you a realistic breakdown.

[Falcon F-14 🇺🇸 @BlueMaxSystems]

@mythos @anthropic @claude For some reason, work on object based strorage schema is suspicious to Anthropic and throws this message in claude code -- "Fable 5's safety measures flagged this message for cybersecurity or biology topics. They may flag safe, normal content
as well. These measures let us bring you Mythos-level capability in other areas sooner, and we're working to refine
them. Switched to Opus 4.8. Send feedback with /feedback or learn more".

Or they are spying and have a keen interest in this topic since they sent a prior prompt asking to use the session for training.. Be alert with what you share!!

[Falcon F-14 🇺🇸 @BlueMaxSystems]

lol! Not…

[Teknium 🪽 @Teknium]

Opus 4.8 the least lazy model ever?

[Falcon F-14 🇺🇸 @BlueMaxSystems]

same experience

[Zae @itzzaeee]

Ok it’s not just me right? Like opus 4.8 seems significantly more dumb and lazy ??? Than previous models

[Infinite Reign @InfiniteReign88]

@ClaudeDevs 4.6 became completely unable to follow a single rational train of thought for the last 3 days… Did nobody notice that? 4.8 is completely unusable always. We need the ones that can still communicate adequately working please.

[Falcon F-14 🇺🇸 @BlueMaxSystems]

Same experience here!

[Falcon F-14 🇺🇸 @BlueMaxSystems]

interesting, yesterday MiniMax M3 was clearly beating opus-4.7 and gpt-5.5 across the board. Late this afternoon it went completely stupid... very odd. Will not follow instructions now - yesterday flawless, today unusable.

Weird.
@opencode

[Mavino Ikein @MavinoIkein]

MiniMax M3 is interesting because the open-weight race is moving beyond “can it chat well?”

The claim is a full agent stack in one model:

- frontier coding and agentic benchmarks
- 1M-token context through MiniMax Sparse Attention
- native image/video input
- desktop-computer operation
- weights + technical report promised after launch

That combination matters more than any single leaderboard row.

If open-weight models start packaging coding, long context, multimodality, and computer use together, the pressure shifts from “which API is smartest?” to “which agent stack can run near my own tools, data, and budget?”

[Falcon F-14 🇺🇸 @BlueMaxSystems]

make sure you use https://t.co/ZXqqUoUNgL and not .com or (cn)

[Falcon F-14 🇺🇸 @BlueMaxSystems]

Instant workspaces!

[dax @thdxr]

@ryanvogel you'll see soon https://t.co/peN2JuHL1W

[Falcon F-14 🇺🇸 @BlueMaxSystems]

High end home inferencing for prosumer…

[Falcon F-14 🇺🇸 @BlueMaxSystems]

DGX is $4,695 now. Comparable to Mac Studio, if you can find one.

[Sprytix @Sprytixl]

https://t.co/qxv5fP5zSD

[Falcon F-14 🇺🇸 @BlueMaxSystems]

@mfishbein i’m

[Falcon F-14 🇺🇸 @BlueMaxSystems]

"Some estimates" from whom, lol... there is nothing in the article saying that compensation is cash based or how much it might be...

[winkle. @w1nklerr]

Nvidia will now pay you to put a mini AI data center on your house

It looks like a normal AC unit in the yard.

But inside sits 16 Nvidia Blackwell GPUs and Dell servers.

A startup called Span builds them, backed by Nvidia.

They bolt onto your home and you get paid for the power and Wi-Fi.

Some estimates put that around $1,000 a month in your pocket.

That is rent money just for hosting a box outside.

Span says it deploys way faster and cheaper than a real data center.

The AI boom is literally moving into the suburbs.

Save this, the grid is getting rebuilt in real time.

[Elon Musk @elonmusk]

😂

[Falcon F-14 🇺🇸 @BlueMaxSystems]

its part of their “substrate”

[Valentin Drăgănescu @escu__valentin]

@Everlier Interestingly enough, "load-bearing", "x is real", "y locked" are also heavily used by GPT... what should we make of this? Is there a way to pin point the training material that resulted in the heavy use of these expressions?