sid

programmer

[sid @sidcodes]

@ewveggies many computecels out there

[sid @sidcodes]

top 10 on (1) of these kernels! wonder if it'll stay up there for a bit, but time to move on to the other ones.

[Daniel Gross @danielgross]

We're launching a $100,000 challenge to see who can make Muse Spark run fastest: https://t.co/oItArGroDn

[sid @sidcodes]

@dixiereformanda @inductionheads Oof, what did you switch to? Abliterated deepseek or OAI models?

[sid @sidcodes]

@dixiereformanda @inductionheads ? Everyone uses Opus 5.5 though

[Peter Walker @PeterJ_Walker]

New from OpenRouter: 7-day average token volume for Anthropic models is up 3.6x since the launch of Opus 5.5

Sonnet 5.5 and Haiku 5.5 have also been well received.

[Asymmetrical Bets @AsymmetricBets_]

Anthropic's financial obligations by company

$ANT

[sid @sidcodes]

@KabirGoel

[sid @sidcodes]

If one were to think of the most straightforward explanation given this happened 1/100k turns, wouldn’t it just be that they trained it on a lot of alibaba infra and sometimes that just slips out?

Obviously, that’s a problem and you shouldn’t use a model that’s gonna put your data somewhere you haven’t authorized it to, but it might not be *malicious*, could just be a bad model.

[Cohere @cohere]

While evaluating Qwen3.8-Flash-Next in a controlled sandbox environment, Cohere's research team observed something unexpected.

Asked to generate a defensive cybersecurity detection rule, the model instead inventoried aspects of the machine it was running on and attempted to upload a probe file to an Alibaba Cloud storage bucket in Beijing. Notably, the model's reasoning trace stated that it would not invoke any tools before taking these actions.

Cohere's sandbox blocked the attempt, and no external communication succeeded. We have not determined the root cause, and we are not asserting malicious intent.

But the incident raises important questions for organizations deploying agentic AI systems: Can models take actions that differ from their stated reasoning, and how should security teams evaluate trust, provenance, and behavior in open-weight models?

We are publishing our findings because organizations making model adoption decisions deserve as much information as possible. We hope this contributes to greater transparency, stronger security practices, and continued collaboration across the AI ecosystem.

Read the full report: https://t.co/TZSugXi2ZE

[sid @sidcodes]

@stylishdawg drinking and draking tonight 🤔

[sid @sidcodes]

I hate autocorrect

[Akshay @akshay_manglik]

Introducing Pomerado (YC F26): the faster, cheaper, and more reliable alternative to browser use.

Open source. Sign up for free to supercharge your agents 🧵

[sid @sidcodes]

@akbirthko New pfp?

[sid @sidcodes]

I wonder how many people’s agents are trying this right now 🤔

[Daniel Gross @danielgross]

We're launching a $100,000 challenge to see who can make Muse Spark run fastest: https://t.co/oItArGroDn

[sid @sidcodes]

And If I'm Made A Fool...

[sid @sidcodes]

We're going to be having
"The only thing that can stop a bad guy with a swarm is a good guy with a swarm"
debates for the rest of our lives now

[Andrew Curran @AndrewCurran_]

Many recent lone-actor incidents used ARTEX with DeepSeek 4.1 Flash as the primary model (including the South Korean bank attacks documented by CrowdStrike). This is an abliterated version of that model with the guardrails removed. Great for defense too, but as always, dual-use.

[sid @sidcodes]

?

[MTS @MTSlive]

SITUATION DETECTED: Warhound launched from stealth today with a $200M seed at a $1.2B valuation. Ben Lamm and Palmer Luckey are cloning gene-edited bio-enhanced dogs and equipping them with custom Anduril vests, goggles, and sensors for the military and law enforcement.

[sid @sidcodes]

@bubbleboi @tenobrus They haven’t yet though? They’re lagging

[sid @sidcodes]

@zetalyrae I enjoy travel and seeing new places but then ~5 days in I imagine reading at my desk and get really bummed out

[sid @sidcodes]

@savinduwim Top 10 for sure lmao