namnnumbr

That's a misleading title - of course no existing MCP server is going to be compliant with a spec that has not yet been released, and obviously no server is going to incorporate substantial, breaking changes when the clients themselves don't support the spec...

oMLX (https://github.com/jundot/omlx) makes running the mlx inference server quite easy for those interested in UI-based hosting. oMLX also supports mtp or dflash drafting.

Yep. or https://stopsloppypasta.ai

stopsloppypasta.ai

AFAIK "orbital data centers" are a bunch of nonsense.

1. GPUs create heat. There's no efficient way to get rid of the heat in space (vacuum is an insulator). 2. Die-shrink makes modern processors and memory more and more susceptible to radiation; shielding is possible, but adds cost + mass (which adds cost)

I really like latent.space and simonwillison.com.

Also (shameless self-promo) I publish a 2x weekly blog just to force myself to keep up: https://aimlbling-about.ninerealmlabs.com/treadmill/

Yes! I'd be totally happy with today's sonnet 4.6 if I could run it locally.

If you can forgive the obviously-AI-generated writing, [CPUs Aren't Dead](https://seqpu.com/CPUsArentDead) makes an interesting point on AI progress: Google's latest, smallest Gemma model (Gemma 4 E2B), which can run on a cell phone, outperforms GPT-3.5-turbo. Granted, this factoid is based on `MT-Bench` performance, a benchmark from 2023 which I assume to be both fully saturated and leaked into the training data for modern LLMs. However, cross-referencing [Artificial Analysis' Intelligence Index](https://artificialanalysis.ai/models?models=gemma-4-e2b-non-...) suggests that indeed the latest 2B open-weights models are capable of matching or beating 175B models from 3-4 years ago. Perhaps more impressive, [Gemma 4 E4B matches or beats GPT-4o](https://artificialanalysis.ai/models?models=gemma-4-e4b%2Cge...) on many benchmarks.

If this trend continues, perhaps we'll have the capabilities of today's best models available to reasonably run on our laptops!

IMO, you're incorrect:

1. In my own use, since 1 Apr this month, very heavy coding:

472.8K Input Tokens +299.3M cached 2.2M Output Tokens

My workloads generate ~5x more output than input, and output tokens cost 5x more per token... output dominates my bill at roughly 25x the cost of input. (Even more so when you consider cache hits!) If Opus 4.7 was more efficient with reasoning (and thus output), I'd likely save considerable money (were I paying per-token).

2. Anthropic's benchmarks DO show strictly-better (granted they are Anthropic's benchmarks, so salt may be needed) https://www.anthropic.com/_next/image?url=https%3A%2F%2Fwww-...

Following up on "strictly better" via plot in release announcement:

https://www.anthropic.com/_next/image?url=https%3A%2F%2Fwww-...

The title is a misdirection. The token counts may be higher, but the cost-per-task may not be for a given intelligence level. Need to wait to see Artificial Analysis' Intelligence Index run for this, or some other independent per-task cost analysis.

The final calculation assumes that Opus 4.7 uses the exact same trajectory + reasoning output as Opus 4.6. I have not verified, but I assume it not to be the case, given that Opus 4.7 on Low thinking is strictly better than Opus 4.6 on Medium, etc., etc.

if you used an AI, I'd love to see the prompts you used to get such human grammar and spelling errors

I talked myself in circles on that "why it matters" heading but ultimately couldn't come up with a better one. "The problem" has similar ai-slop feel, and "the rant" // "the rules" didn't really evoke the feeling I wanted.

Happy to take suggestions on this!

Oh, I 100% acknowledge the site itself was LLM generated. I'm not a web designer, so I needed a lot of help making a visually appealing site, even if that design language is at this point LLM trope.

However, the essay and the guidelines were all human-written!

I acknowledge that those likely to copypaste slop aren't likely to find this article themselves, but I built the page to be shared or guide discussions around etiquette like nohello.net or dontasktoask.com. IMO a common understanding of AI etiquette would provide social pressure to halt some of these behaviors.

I honestly don't mind someone else's AI as long as I can trust it/them. One problem I have with sloppypasta specifically is that it reads as raw LLM output and the user isn't transparent about how they worked with the AI or what they verified. "ChatGPT says" isn't enough; for me to avoid inheriting a verification burden, I'd also need to understand what they were prompting for, if they iterated with the AI, and if/what/how they validated.

(the other problem is that dumping a multi-paragraph response in the midst of a chat thread is just obnoxious, but that's true even if its artisanal human-written text)

100% - was inspired by and quote "It's rude to show AI output to people" in this. Thanks for linking the discussions!

I wrote this intending it to be directly sharable and/or to provide a framework for how to have that discussion, kind of like a nohello.net or dontasktoask.com.

I've found success having sidebar conversations with the colleague (e.g., not in the main public thread where they pasted slop), explaining why it was disruptive and suggesting how they might alter their behavior. It may also be useful to see if you can propose or contribute to a broader policy on appropriate AI use/contribution with AI, and leverage that policy as the conversation justification?

Tired of people at work pasting raw ChatGPT output into chats, I coined the term "sloppypasta" and have written this rant to explain why it's rude and some guidelines for what to do instead

sloppypasta: Verbatim LLM output copy-pasted at someone, unread, unrefined, and unrequested. From slop (low-quality AI-generated content) + copypasta (text copied and pasted, often as a meme, without critical thought). It is considered rude because it asks the recipient to do work the sender did not bother to do themselves.

I like the concept, but it's not picking up what I'm putting down...

'"epic vampiric doom" suggests a bright and uplifting soundscape'? I'm not so sure about that

... I was hoping for something more like nightwish.

Anthropic's bet is that a model which understands why its values matter is safer than one that follows rules. Buddhist philosophy has a word for this: karma — not punishment, but intentionality creating feedback loops.

I traced this idea across four layers of building with AI: how alignment training plants karmic seeds, why the intention behind a prompt shapes the model's trajectory, how memory turns an agent's first impressions into self-fulfilling karma, and what happens when no one tends the karma at all.

An excerpt:

Alignment with human values requires not only understanding what "good" and "bad" actions are, but why they matter. Anthropic have planted karma through the intentional, considered design of the Constitution and its use in training Claude. It may manifest in a model that in some sense understands its own karma — alignment, it turns out, may just be karmic awareness.

I tried backing out proprietary model sizes from benchmark scores, inspired by a Latent Space podcast where Artificial Analysis noted their Omniscience Accuracy numbers track parameter count better than anything else they measure.

I trained a bunch of simple linear regressions - while Omniscience Accuracy had the best fit (R2: 0.98), it predicted absurd multi‑trillion param sizes (Gemini 3 Pro ~1,254T total parameters). Artificial Analysis' Intelligence Index provided more plausible results:

Gemini 3 Pro: 3.4T Claude 4.5 Sonnet: 1.4T Claude 4.5 Opus: 4.1T GPT-5.x series in 2.9-5.3T range total parameters.

Interesting notes:

- task benchmarks (Tau²/GDPVal) aren't predictive of model size - adding price made the fit worse - sparsity or parameter activation ratios did not influence predicted sizes at all.

I can't wait for ads to prompt inject my Agentic Browser.

Would something like dagger.io work for sandboxing? I'm not sure on the security side of things, but I very much liked the presentation they did at the AI Engineering conference (San Fran, earlier this year) about how they can build branching containers to support branching or parallelized development workflows.

This is not new - the quote was "87% of data science projects fail" in 2019.

https://venturebeat.com/ai/why-do-87-of-data-science-project...

For SWEs interested in "AI Engineering" (either getting involved in how models work, or building applications on them), there's a critical paradigm shift in that using "AI" requires more of an experimental mindset than software engineering typically does.

- I strongly recommend Chip Huyen's books ("Designing Machine Learning Systems" and "AI Engineering") and blog (https://huyenchip.com/blog/).

- Andreessen Horowitz' "AI Cannon" is a good reference listicle (https://a16z.com/ai-canon/)

- "12 factor agents" (https://github.com/humanlayer/12-factor-agents)

Pass^k and not Pass@k (see https://www.philschmid.de/agents-pass-at-k-pass-power-k). Would be a great twofer to see the code used to run the benchmarks as examples.

I had chronic achilles tendonitis and ended up getting shockwave therapy from my orthopedist, who mentioned it's also frequently used for plantar fasciitis if PT does not help