Thinking Avocet

@tenobrus I think it is a mistake to think of this as a game theory move I think the Chinese administration is just techno positive

people are saying a lot of things in response to this that still don't make sense to me - "commoditize your compliment" that only works if you are in fact capturing the value, as far as i can tell it doesn't meaningfully apply to models. moonshot as an inference provider for kimi is undifferentiated or even worse than others running exactly the same model. open source labs capture none of the value - "xi doesn't believe in ai risk / hasn't read hpmor" sure i mean i didn't fucking think he did lol. but there is clear already proven cyber warfare capability, and beyond that if a technology is an accelerant to ur country then you don't want competitors to have access to it purely on very normal tech grounds. u don't have to worry about nanobot xrisk to not sell ur latest fighter jet tech to other militaries. - "this is what china does, they sell cheap knockoffs to take economic power away from the americans" but in fact they aren't selling anything, they're giving away an artifact that lets everyone else reap the gains. maybe it's true that this *reduces lab valuations*, and generally i can absolutely see the angle that it *reduces american differentiated power* since the frontier american labs can no longer deny frontier intelligence to non american companies.... but china would benefit much more here by *being the only other country* who also has this power rather than fully leveling the playing field . - "something something soft power and image" yeah i mean maybe. it's not that i can't see it. it's just tough to see how image meaningfully changes if new very powerful models are locked behind an api? what a few ai devs on twitter like china better? doesn't seem like the kind of thing xi cares about, and if he does it seems clear it would be outweighed pretty fast once the concrete economic benefits become big enough - "it introduces chinese ecosystems and ideologies around the globe" idk man like maybe but only if they've added some really well hidden secret triggers yknow? k3 still claims to be claude and has basically claude's cluster of western opinions . so the claim here would have to be secret sabotage, and even that is just not that hard for someone to eventually detect and then publicly announce plus tune out

@deepfates @catboosted

@catboosted oh cool Claude raises our kids now

This is honestly really fucked up

People wanted a timelapse so I asked Codex for it and this is the best it could do

@Angaisb_ yeah, I feel like the old one had a nice cuteness and definitely see that the timelapse is probably easier to get on a new run. It would be fun though! (And an excuse for another post)

@thinkingavocet the first one took like 10-15 minutes, this one took over 5h 52m, using 2,627,274 tracked tokens the models in this one look much better, but I agree lighting is better in the first one it's going to be hard to show a timelapse over all the work it did but I can try

GPT-5.6 Sol Max after working for 6 hours

In fact, this same process is happening today

@emollick lmao

Fable, turn my tweet into a thinkpiece (this was pretty funny): There has never been a better time to have opinions about artificial intelligence. I say this with some authority, because I am currently having one. Consider the perfect storm we find ourselves in. On one side, AI has become the ideal subject for speculation: consequential enough to matter, ambiguous enough that no one can be definitively wrong, and moving fast enough that last month's take is safely forgotten before it can be checked against reality. On the other side, AI has become the ideal producer of speculation: tireless, fluent, and capable of generating a thousand words on "What GPT-5 Means for Democracy" in the time it takes a human columnist to open a blank document and feel bad about it. We have, in other words, achieved thinkpiece singularity. The subject and the instrument have merged. The snake is not merely eating its tail; it is publishing a 1,200-word reflection on what tail-eating means for the future of snakes. The Economics of Infinite Speculation The traditional thinkpiece economy ran on scarcity. There were only so many columnists, only so many column inches, only so many hours in which a professionally opinionated person could convert anxiety into prose. This scarcity imposed a kind of quality floor — not a high one, but a floor. Someone had to care enough to write the thing. AI removes the floor. The marginal cost of a take is now approximately zero, which means the supply curve for opinions about AI has gone vertical at the exact moment demand for opinions about AI has also gone vertical. Economists would call this a market clearing at infinite volume. The rest of us call it LinkedIn. But here is the twist that makes this era genuinely strange rather than merely noisy: AI is uniquely suited to writing about AI because AI writing about AI is the one genre where the author has genuine subject-matter proximity. When a language model speculates about what language models mean, it is, in some sense, the only writer with firsthand experience. The thinkpiece about AI written by AI is simultaneously the most derivative and the most authentic document of our time. No one knows what to do with this fact, so we mostly don't mention it. The Four Immortal Takes Every AI thinkpiece, human or machine authored, converges on one of four positions, which I will now enumerate in the traditional listicle format that signals seriousness: It changes everything. (Bullish, well-compensated, frequently keynoting.) It changes nothing. (Contrarian, tenured, secretly using it to grade papers.) It changes everything, and that's bad. (Prophetic, exhausted, right about something eventually.) The real change is the discourse itself. (Meta, insufferable, this essay.) The remarkable thing is that all four positions can be argued persuasively with the same evidence, which is why the genre is infinitely renewable. A benchmark result, a chatbot transcript, a layoff announcement — each is a Rorschach blot that faithfully returns whatever thesis you brought to it. AI writing tools have simply industrialized the Rorschach reading. You can now generate all four takes from the same news event before breakfast and A/B test which one performs. What Gets Lost It would be easy to end here, on the ironic note, because irony is the load-bearing wall of the genre. But the honest version of this essay has to admit something less comfortable: the flood is drowning something real. Buried in the infinite speculation is a small amount of actual thinking — people running experiments, reading papers, talking to workers whose jobs are changing, noticing things before they become narratives. That work was always rare. It is now rare and camouflaged, indistinguishable at a glance from its synthetic imitators, because the imitators have gotten very good at the surface features of insight: the confident hedge, the counterintuitive framing, the paragraph that begins "But here's the twist." (You may have noticed one of those a few paragraphs ago. The camouflage works in both directions.) The old signals of effort — length, fluency, structure, the presence of a numbered list — no longer signal effort. This is a genuine epistemic problem, and it will not be solved by detection tools or disclosure norms, because the issue was never really authorship. A lazy human take and a fluent machine take are equally worthless; a rigorous piece of thinking is valuable regardless of how many silicon collaborators touched it. The question was always whether anyone checked anything, and that question was hard to answer before the machines showed up. The Useful Version of the Golden Age So here is the optimistic reading, offered with the mandatory caveat that optimistic readings are position #1 and I have already mocked position #1. Golden ages of cheap production have historically been terrible for average quality and excellent for the tails. The printing press produced mountains of pamphlet garbage and also the scientific revolution. The blogosphere produced infinite noise and also a handful of writers who could never have gotten past a magazine's gatekeepers. The pattern is consistent: when publishing costs collapse, the middle of the distribution rots and the extremes flourish. If that pattern holds, the AI thinkpiece deluge is not the end of thinking about AI — it is the end of mediocre thinking about AI having any value at all. The take that could have been generated in eight seconds is now worth exactly what it costs to generate. What remains scarce, and therefore newly valuable, is everything a language model cannot do on its own: the experiment actually run, the data actually gathered, the prediction actually registered and later scored, the sentence that could only have been written by someone who was in the room. The Golden Age of thinkpieces, in other words, may turn out to be a gold rush in the classic sense: most of what glitters is pyrite, the real money is in selling shovels, and the durable fortunes go to whoever figures out what the territory is actually worth once the prospectors move on. Until then, the speculation will continue, at scale, about scale. Somewhere right now, an AI is drafting a response to this essay, and another AI is summarizing that response for a newsletter, and a human is skimming the summary and forming an opinion they will describe as their own. It is the most efficient opinion supply chain ever constructed. Whether anything is being thought is, appropriately, a matter of speculation.

Between having AI as a subject for infinite speculation and using AI as a writer to produce such speculation at scale, it is the Golden Age of thinkpieces.

Fast mode isn't worth it. While token speed is faster, a good chunk of time goes on running of deterministic tests, hence your effective speedup is much smaller than advertised for 2x+ the cost. If you want to go faster, tell your agent to validate less, it would probably get it wrong more, but not as often as you think. They often go around in circles validating what's working fine already.

@_xjdr violation of object permanence ?

i get attached to my tmux windows and their location and order and when they move (for various reasons) it ruins everything and i get upset. im pretty sure there is a name for this ...

@ChaseBrowe32432 this guy seems to pull it off

Is there anyone who can steelman the “tell your customers they’re losing their favorite product in 5 days, then say sike over and over” strategy

@jbulltard1 @argupta x.com/thinkingavocet…

@argupta They can’t just keep doing 1 week delays

What exactly is the gimmick here? They can’t go back to opus after this and they can’t subsidize fable at what they’re charging? They can’t keep doing this indefinitely

The year is 20XX. Vibe coders are resting, for today is Sol-day [1]. Ceremonial Anthropic speakers dotted across the city announce: “We're extending access to our most powerful model (Fable Ω.5) to all paid plans, until next week, April 6.” Zoomers that are still on the $20/month plan breathe a sigh of relief. [1] officially renamed in honor of the first Fable competitor in the Treaty of Europa, which ended the Second Tensor Hyperwar

@irl_danB it's primarily because of foom-ers right? thinking that recursively improving AGI discovered = full molecular control of the universe imminently (I personally disagree with this)

not only is this possible and likely, it’s the only world that’s ever existed very high bar to show that somehow AI introduces a type of unipolarity that has never existed before. and for some reason many people start with this as some sort of default premise

@TheZvi fable is the first model where i'll ask it to do something and despite not even so much as hinting as to why i'm asking for it it'll correctly clock the *ultimate* goal of the project even though that requires quite a few logical leaps

@hackerdocc this happens in the native harnesses as well btw (fable has decided to only use python to edit in a long-running convo for me ... hooray for token efficiency)

honestly not sure how i feel about custom-harnesses anymore man, everything i use clearly has more errors than the native tools the models are trained with. my gpt 5.5 just wrote perl scripts cuz their previous 3 attempts of using the edit tool in pi failed

@tszzl I've wondered this, but the most annoying verbal quirks are encountered specifically when they are reward hacking you socially (making excuses / downplaying issues). Personally, I just want them to say "uhhh shoot yeah, that's right, my b"

hypothesis: the writing styles of language models are basically fine, they weren’t better in some halcyon before times. we just use them so much that we get annoyed by their mannerisms. they need to have a superhumanly diverse idiolect to not become grating

@amorriscode @HamelHusain the toggle almost never works for me.. i use forking a ton, which might be relevant

@HamelHusain in the desktop app it's a toggle in settings. will look into the effort issue.

You can have always on remote-control for claude code by enabling this in the config ~/.claude/settings.json { "remoteControlAtStartup": true, ... } Make sure your'e logged in with subscription. It's can be brittle, for example when I change /effort on the host it hangs on the client. But, its a huge quality of life upgrade

The "should you read code" debate is dumb because the real decision isn't binary, it's a scale: 1. Reading every line of every diff 2. Scanning every diff, reviewing important lines 3. Ignoring diffs but understanding the 'why' of every PR 4. Spot checking PR's instead of reading every one 5. Ignoring PR's, but doing regular spot checks on the codebase 6. Ignoring the code, but spot checking agent traces to help improve the system 7. Ignoring both the code and the system, let models handle everything Where are you on the scale?

@dejavucoder @yacineMTB c a c h e

@yacineMTB name 5 reasons

Subagents are an antipattern

@rezoundous it's a token saving technique for when you are doing something that involves a lot of output tokens, but don't need the taste of Fable when writing I don't expect it to be generally applicable

My experience: Fable 5 (orchestrator) + GPT-5.5 (executor) is better than pure GPT-5.5. But only by a little. What am I doing wrong guys.