serf

>How many tokens/s do you need for a coding agent to feel like Claude or codex?

it's kind of a loaded question, this model gets a lot of tok/s but reasoning is forced on; the time-to-answer is really a combination of things like "how much does this thing want to over-reason" and "how much does this thing like tool use?" .

i'm gettin between 200-320 tok/s on a 5070ti with 131k context on llamacpp and no tuning.

very nice.

I can't speak to the intelligence of it yet.

these small agentic-focused models are a lot of fun to play with if you haven't yet.

I don't think fast gpu rendering terminals should be mandatory for this kind of thing, and if you should have an option that facilitates the need for one then might I suggest falling back to something sensible rather than stalling out for forever when the user doesn't render glyphs as fast as you want them to?

either that or fail outright; dont waste my time.

the reality is that evil maid attacks, and more broadly (and realistically) physical access attacks are exceedingly rare. The vast majority of stolen credentials are phished or some other social-engineering trick, not James Bond sneaking in with his badusb.

the picture galleries on that site are great, I wish I had chronicled things that way.

having used both I think bonsai 2 is far ahead, but it still calls tools too often. This isn't always a bad thing, but it seems to have more tool calls per successful run than other equivalents; like it's in-grained to triple check things.

might be a good characteristic for a quantized model to have ..

>Obviously you can prevent that from occurring with some extra infrastructure in-place, but at that point you're basically creating source control.

most of the harnesses are checkpointing most of the data of a git commit aside from the formal release and messaging, anyway; thats why they all have a similar problem with SQL dbs exploding in size after a bit of time.

Right now we're in this weird transition where most of the harness systems are layering git tool use w/ the constant snapshotting AND a memory subsystem often, probably leaves a lot on the table in saving wasted token use to just ditch the vcs all together and rely on the already persistent and existent checkpoints and memories for reconstruction.

>Needs LLMs to encrypt a flash card.

throwing a prompt at an agent is often faster than cracking out man, lsusb, lsblk and friends, doesn't mean he can't do it without.

if I send a prompt to make an assortment of alpha numeric directories or whatever other braindead-to-automate-with-wildcards-and-ranges thing it doesn't mean I don't know how to do it, it means i'd rather think in English than regex and bourne shell and the extra few seconds and the token spend is worth the luxury of avoiding them for me, at that time, for that task.

prediction market betting for the brave and obvious.

i'm pretty convinced that if you look at the cadence and tempo of updates on Steam among actively developed games that one could probably make pretty good guesses about which groups are using LLMs internally.

Doesn't prove anything, lots of stuff changes development house tempo, but to anyone watching things as a whole updates across steam games in active development are 1) quickening 2) narrowing 3) becoming predictably scheduled ; all of these point to more than just a single group refocusing on the work at hand.

I think builds are getting pushed out a lot quicker, and personally I know of a lot of anecdotes where groups' entire build and output chain are handled by LLMs invoked by cron (or a clone). That alone will add a lot of regularness to things.

"the audacity of these people, they didn't even mention that they cheated and used emacs ! they sacrificed their pinkies for development time!"

I tested it with over 200 mods and a save from 1.6 and it worked.

the client bitched about the mods having incorrect version numbers, but after re-enabling them and loading the save nothing was broken (any more than it was in 1.6, at least.) -- and it loaded much faster.

I think there was a lot of in-situ handholding for unity API expectations.

>What would be your definition of not running?

inflation combined with little to no wage increase leading to total economic stall outside of the scope of any safety mechanisms, which precipitates an extreme cliff-face degradation of everyday life and commerce for citizens at every level.

or, another way : conditions that would lead to public rebellion en masse, which further stalls things past the economic strife.

the type of thing 1780s France saw.

>Dignified I don’t think it ever was. What do you think made it that?

it was a car designed to get rural farmers to stop riding horses everywhere.

it's slightly more dignified when your mode of transportation doesn't take a literal shit in the street.

>.. nor any sort of agentic work.

may I ask why?

oai luna runs at 60-70 toks/s; astra runs around 35-45 , are you just worried that your agentic work can't spread across a lot of parallel work?

it's okay.

it's minimal and somewhat opinionated, not integrated well with memory subsystems (although that's kind of the point)

my only real criticism is that last I played with it it deals with slow provider responses very poorly , so lots of locally hosted llms require a lot of tuning to get it to handle streaming/non-streaming/timers well.

way less minimal (but still pretty minimal) I prefer Pi; but different markets I think.

hax fit my openbsd laptop well, espescially since Pi and related derivatives are kind of un-spokenly incompatible with the BSDs.

"but my vibes-based opinion from using it is that it's basically on par with Claude Opus 5.5 for software engineering." is an outright lie that's qualified by a "well in my opinion.."

unless you're comparing the most extreme specific harnesses to claude code that's just false.

..and I have a lot of disdain toward Anthropic , & I think GLM is great (although too expensive outside the z plans), but it's just wrong by every scorable metric.

spinning up a GPU and as many cores as possible?

mousewheel repurposed?

background follow your mouse?

neutral tones?

well then YOU are viewing a frontier model announcement!

>Sure, but is that a negative for the stock price?

I feel like that statement is one of mans' biggest foot-guns in the modern world.

that single sentence has driven the biggest hucksters the world has known to commit the biggest frauds ever perpetrated against man for decades.

yeah but since everything in our subjective experience affects the brain, is it a point worth discussion?

"Can singing 'Twinkle Twinkle' while on my commute affect my brain?"

(yes, and in fact so did asking that question, receiving an answer, and reading this sentence, even this paren -> )

every single 'normie' i've seen use Discord can barely figure it out, they chase red badges across the UI all day and really have no clue about the paradigm at all.

They're there of course because some 'normie' thing of theres inevitably requires the sign-up of a discord for bonus game skins or some other such stupid shit, but the fact remains that they're clueless.

much like the average Outlook user, they're there because they were dragged into it by another element in their life (their job) and they really have no background or clue on how to use the piece of software, they just know they need to pretend and cite experience on the CV for a job.

"On February 4, 2026, following the restriction and closure since 2025 of other government-funded information websites, the CIA announced, with no warning and no explanation, that The World Factbook was discontinued."[0]

[0]: https://en.wikipedia.org/wiki/The_World_Factbook

>We need to find better ways to sustain great journalism while making it easy for people to pay and consume.

journalists have always been employed and funded by the rich and powerful so that they can 1) amplify good news about the rich owner 2) drown out bad news about the rich owner 3) push agendas 4) resist opposing agendas.

the concept that at one time journalists were some independent rag tag group of truth-seekers is blown away by any historical recollection of the practice.

so : no , I don't think that we need to find better ways for people to pay and consume ; I think that's entirely up to the profiteers behind the mechanism to figure out how to do.

It's never been easier in history to be a nobody and get a message out to the masses; this is one of the only times in history where a journalist can do their job without selling their souls to the devil (paraphrasing hunter thompson).

and this is ignoring recent internet history of ad-malware, person-tracking, and compulsory clickbait headlines that beg the stupidest low-IQ takes possible in order to get you to click into a paywall.

I feel very little sympathy towards industries that'll save themselves from drowning by stepping onto their customers' heads to avoid the ocean.

'The Road to Wellville' is an interesting comedy movie regarding Kelloggs' health retreats and all this stuff.

if you're into yakuza campyness try the outrage series. he's great in them.

most of that is trivial to get around by claiming permission. really.

"X license disallows this current use without express permission"

"Z gave me permission , here's the pdf, they gave us permission regardless of what licensing says going forward."

<insert pdf generated from a text file that says "I'm Z and I gave permission.">

"Okay, we will continue forward while noting that Z gave us permission regardless of what licensing says going forward."

isn't the person that did the tron typography visuals a regular here?

cool to see it come full circle and see another person thrown into the industry by the work of another forum member.

now add a kernel level rootkit and a monthly subscription which guarantees a set amount of loot-boxes a month.

i've read enough 'shlemiel the painter' ports across hundreds of systems and thousands of examples that I must quickly and coldly dismiss the precept that computers were ever used efficiently.

they got rich by wanting to get rich and make money; that's the entire motivation and it doesn't have a ceiling.

you don't just get rich and stop, it doesn't stop for these people. They aren't rich enough. They aren't powerful enough.

Most folks at that table with Trump are way beyond the joy of creation; they're playing politics now, and their game is the competition with the 200 other hyper-elite for positional vying.