Terretta

as often the case when Apple makes its move, their implementations -- while late compared to e.g. Garmin which has had them for years -- purport to be leading accuracy on these features among wearables

Self-selected ignorance and scrambling seem both correlated and causative.

That said, SOTA machines can't run an office snack bar yet, so it'll be a minute before abundances other than caution.

Except the "one model" is too small for a 64GB Mac much less 128GB, sad since the Q3 is proven less competent.

Offering a Q boost (with no leave behinds) on first run would be a bump worth some vibe coding while.

TY for link.

Unless trying to use it interactively and adversarially, in which case it's not fast enough plus would be why those of us without our own datacenters will get told we can't have nice things.

On the contrary, the readme for this project, `litelm`, is contrasting with the bloat project, `litellm`. This project subty deboats an L.

The bloat (or featuritis) project has the 700mb and 100k+ LOC, this de-bloat has 2k LOC and specifies its only two deps ones most people want as deps (although httpx should be swapped to its successor).

So this response is off base, the 700Mb isn't forgotten, it's been dropped.

"Congress gripped by AI panic after doomsday warnings"

https://www.axios.com/2026/09/11/congress-ai-anthropic-coxon...

Not relaying to us simple folk. The message is for government simple folk, amplified and relayed by us simple folk as constituents.

The common element is lobbying, for regulatory capture, to help pull up the ladder.

They don't have to be colluding, they just have to hear the same things from the gov at the same time (as they would), then it goes in media waves beacuse journalists don't as easily get published for a story about one thing as they can if two or more examples make a pattern.

You also don't need any latency in the loop or moving parts that could wrongly interfere with the reliability of making that money.

Fortunately, the last thing this is is "opaque". You have all the observability you need to do what you want to do with unexpected usage.

What AWS (and every CSP) gives you is all the instrumentation and out of band tools to implement your own angel of death, to API kill a runaway thing if that's what you want to do.

You can also find countless TF and CDK patterns for it if you aren't sure how.

But please, no, do not ask to make critical systems subject to the whims of some billing decision enforcer.

The "why did they cut me off" complaints here far outnumber the "runaway bill" complaints.

https://www.cablematters.com/pc-1640-180-intel-certified-80g...

aka: https://www.amazon.com/Cable-Matters-Thunderbolt-Black-3-3ft...

But we prefer such be braided, so with CalDigit we use this:

https://www.caldigit.com/thunderbolt-5-pro-cable/

https://www.amazon.com/CalDigit-Thunderbolt-5-Pro-Cable/dp/B...

Or with Anker we use this:

https://www.amazon.com/Anker-Thunderbolt-Charging-Bandwidth-...

That said, we mostly use Apple TB5 for fixed work stations where cables don't get lost.

https://www.apple.com/shop/product/mdw94am/a/thunderbolt-5-u...

We avoid gaming brand TB cables and the scrabble letter brands.

Anker offers lovely soft silicone sheathed USB-C cables, in black or white and many lengths. They perform for data and do not get stiff or discolor. Our office specifies these unless TB4+ is required.

Might be visually cooler and a gesture at decayed compute history if the decay pixels land as black and white (1 and 0) instead of color.

RBG noise damage does happen digitally for image damage, while hue protected noise damage is more organic -- a more filmish look. Most people prefer hue protected noise (saturation and lightness variance fine). It's likely this would look older and more decayed given our biased association with film grain.

For this, monochrome noise with a target grey level (e.g. 20% grey) would be visually nice, but the 1s and 0s would be even more old timey computer museum decay.

Fast follower persona clusters around emerging zeitgeist across the tellings. At the moment, arguably that's mostly Qwen for everyday hobbyists, and GLM for those that can run 512GB to 1.5TB of memory. This persona is seeking viable applied results: "I have frontier at home".

The early majority pick things up after models are curated into apps like LM Studio or one's platform app of choice, usually at least one major release behind because it takes that long to choose and package into mass distribution.

This is the step where early majority persona "has no idea" what the parade of weird names is about, they care about qualia of the conversations they try to have.

This persona is, at present, very under-served, and likely to remain so until mass devices can perform feeling like 27B at Q4 large quality better, or workplace devices can achieve a pragmatic utility like 135B at Q8 or better.

Harnesses that work where the workplace persona lives bridge this. This persona doesn't care the Chinese model name, they care "does it code?" For that, the applied harness and model take time to be matched, as JetBrains did harnessing a tailored Qwen 3.6 in the IDE. More efforts like https://www.jetbrains.com/junie/ are needed for the majority persona to perceive value from changing their workflow again.

HN's "job" is better outcomes with less friction at each persona.

It's curious how much of these could apply to either:

https://en.wikipedia.org/wiki/Reconstructive_memory

https://en.wikipedia.org/wiki/Misinformation_effect

And many mechanisms exacerbate issues by entrenching preexisting issues.

A Macbook Pro M4 or M5 drives TWO of these at once, well, for equivalent of quad 4K configured as dual super ultrawides:

https://www.samsung.com/us/monitors/gaming/57-inch-odyssey-n...

Per Apple support page, MBP with M5 Max handles:

Two external displays

Two displays up to a native resolution of 8K (7680 x 4320) at 60Hz, 5K (5120 x 2880) at 120Hz, or 4K (3840 x 2160) at 240Hz over Thunderbolt or HDMI

Three external displays

Two displays up to a native resolution of 6K (6144 x 3456) at 60Hz or 4K (3840 x 2160) at 144Hz; plus one display up to a native resolution of 8K (7680 x 4320) at 60Hz, 5K (5120 x 2880) at 120Hz or 4K (3840 x 2160) at 240Hz over Thunderbolt or HDMI

Four external displays

Four displays up to a native resolution of 6K (6144 x 3456) at 60Hz or 4K (3840 x 2160) at 144Hz over Thunderbolt or HDMI

https://support.apple.com/en-us/101571

PS. Use TB4 or TB5 dock from Cal-Digit (match the TB version of your mac), or this or one of the rebranded versions of the same OEM: https://www.owc.com/solutions/thunderbolt-dock, or an Ivanky Fusion dock if you want to go nuts with up to SIX displays, Quad 6K@60Hz and Dual 4K@60Hz:

https://www.amazon.com/Thunderbolt-Monitor-Docking-Station-A...

Just Tailscale into the Studio from iPad Pro 13" with magic keyboard and Kit Knox's rootshell:

https://github.com/kitknox/rootshell

Note that the iPad Pro can also drive a 4K second screen if you like, and most anything else a Macbook with a single port could drive.

Why not Macbook Air? Because the iPad Pro is also a tablet, touch screen, and 5G...

Someone's pockets, must be gnomes:

• PHASE 1: COLLECT UNDERPANTS. (tariff "foreign" goods desired by Americans, collect upon import from companies, clarify spiked prices experienced by consumers ‘entirely’ due to prior admin policies)

• PHASE 2: … (let simmer)

• PHASE 3: PROFIT. ("return" tariff costs to companies)

Principle 1: Be aware of your own anthropomorphic cognitive biases.

Principle 2: Be skeptical of others' (and your own) hypotheses. Design control experiments for possible alternate strategies that could produce the observed behavior.

Principle 3: Design novel variations of stimuli or benchmark items to test robustness and generalization.

Principle 4: Be curious about mechanisms underlying performance.

Principle 5: Consider performance versus competence.

Principle 6: Analyze failure types, and embrace “negative” results.

Transcript, with podcast audio for download, Apple, Spotify, or direct stream.

anyway, real feedback:

this is one of the longest existing unix service patterns ... what's this doing that the distro and usual suspects don't? (as a startup ISP in early 90's, we found what we needed already in BSD 3 distro)

to browse countless variants, swing by github, search terms like angel, daemon, dead man monitor, etc., although to be fair, most of those will long predate vibe coding, so this one may bring properties those don't

such a tool likely shouldn't be yet another separate SaaS, and the 'narrower' point is why: it should not feed that separate Saas (opposite of narrow), it should feed OTel, PagerDuty, whatever is your "broken things" sink, dash, and comms tree sender ... if you don't have that, do that first, then do this class of thing

* Results useful as they differ from pop theory, but mostly clustered enough you'd pick the language for other reasons.

https://www.noodlesoft.com/

It does lots more, but this bonus is great.

Whether selecting for ads or content, one had to figure different ways of getting after it, whitelisting was already off the table.

We'd found "structure" could work, but had to be fuzzy and adaptive as marketing shuffles pages around constantly, not just on the whole, but for A/B runs and personalized "experiences".

So that seems back to, hey, models could help.

(Though none of them, to date, attain as high success as latest hand coded "reader" tools, worth inspecting for how this could be generalized.)

Try to not get more upset when readers coach you than when the LLM rewrites you. Both happen, both are fine.

STRONG agree. Most devs, if they can see their costs, try. Even just putting it in a starship prompt goes a long way!

Observability supports effective, responsible, and accountable use, "cost controls" (as typically enterprise defined) supports bean counters. Ensuring Observability is recognized as a valid Control™ should be on every engineering executive's TODO list.

> the cache is the dominating cost reduction force

TY for the rich reply. As you're very aware of the cache issue, it might have been nice to have the word "cache" more prominent on your "Key Techniques" image (currently Lower Cost Models, Smart Routing, Spend Controls, Context Optimization; cache tuning buried at the end) or higher in the body sections.

It was mostly the image that prompted (ha) me to call out the marketing driven angle.