Let's face it. Claude (in particular) is a terrible writer. There's a whole cottage industry of skills and CLAUDE.md instructions trying to push it toward writing better, but each new model iteration seems expressly designed to override all that so that it can load up its writing with unnecessary participle phrases, not-this-but-thats, burying the lede, and other nonsense.
I genuinely wonder if the people inside Anthropic actually communicate with each other like that. Has it been imprinted with Dario's engrams?
After some iteration, this is what I currently shove in every AGENTS.md to make Fable legible:
## Writing guidelines
These apply to documentation, code comments, commit and PR messages, and replies to the user.
- Write precisely in clear, complete sentences; keep text concise and proportional to task complexity.
- Stay focused: avoid filler, repetition, over-the-top detail, and tangents the user did not ask for. Once a fact is stated, do not restate it for effect ("so the commit landed on a branch nobody was going to merge"). Do not editorialise.
- Always prefer ISO 24495-1:2023 conformant plain language over dense technical jargon: short sentences, one idea per sentence, define terms on first use.
- When reporting your own mistake, give the cause and the fix in one sentence each; no apology, no framing ("the mistake was mine"), no post-mortem.
- Never use em dashes or cataphoric teasers such as "Here's the thing" or "But there's a catch".
This seems solid. It is so frustrating when you state a fact and it goes off and does research and confirms that indeed the fact that you stated is accurate.
I was incredibly surprised how much better was Astra at rephrasing the docs that Fable came up for a project I am about to publish. The instructions are now actually what a human would expect, with clear logic flow diagrams, lists itemized, short paragraphs. It also actually correctly detected the train-of-thought leftovers, as well as stuff that never made it to final version of the project and removed them.
Meanwhile Fable consistently ignores all my requests to write this exact way. I mean, the bare minimum I ask it for it to itemize lists and not write in single long passages using comas, semicolons and 'and's. Still ignores them.
I honestly think it's time to call Astra the SOTA. It may not lead all the benchmarks but it genuinely feels much superior of a model. Not to mention the ¢20 Codex plan with frequent resets (https://codex-resets.com/) gives me roughly as much allowance as the ¢90 Claude plan, especially with recent limit cuts on Anthropic plans.
Maybe check in the other direction as well: Astra to Fable.
I frequently simply let one of the three review what something that looks like awesome output by one AI gets totally annihilated by the other.
Finished outputs are easier to improve than bend a LLM to produce stuff like that in my observation.
Same with Gemini.
I yet have to find out how to handle this, whether I let agents check themselves and if on what process step.
Tweaking is hard.
I agree with your conclusion I am a huge ChatGPT and Codex fan, Gemini has to many infrequent quality changes when new models arrive ranging from great improvement to WTF.
ChatGPT seems to get scaling well while Claude still feels unstable, unclear usage statistics. Really weird.
These resets also reset your weekly timer right. So it’s like, you may have 30% left for 1 days that you want to use. They reset it, and that means you your “new week” timer starts today.
That sucks, because it doesn’t always work in your favour if you plan your weekly spend.
I think a real reset shouldn’t also reset your week timer.
Yes, I was saving my 95% usage for the last day. Then they reset it. Now I have lost 95% usage!! Im back to 7 days and got locked in at only 5% usage for the week - that's actually theft.
Agreed. But they also hand out the usage resets you can use at will, so if you bank couple of those, they come in handy specifically in situations like you described.
It reminds me of when people will use filler words because they haven't actually decided what thesis they're going to commit to yet. Just-in-time thinking.
Not just imprinted with the engrams of any one person (however brilliant they might be) — it’s more than that! Like all great writing, it’s more like it’s been imbued with patterns, patterns that are peppered with a potpourri of popular ways of writing in an enthusiastic style while regularly restating things and recapitulating them. Would you like me to explain more?
I love that description. It matches very well what I feel about LLM output. This overly sweet scented smell meant to cover up something nasty, with a strong artificial tint to it.
Which doesn't say that much about the LLM itself but about the people that make the training material.
Do you suppose it's just that it's aping humans who say too much without saying anything or do you think maybe this is a stalling tactic to get people to spread out their transactions?
Like how theme parks generally don't do much to keep queues short (or Disney charges you a premium to skip the queue)
> Do you suppose it's just that it's aping humans who say too much without saying anything
Considering how much of the input must me nonsense SEO bullshit articles and blogs that only serve to promote a person or company that might be a factor.
I also often have wondered if it is also targeting those same people. Certainly with tools like deep research options (not just Anthropic's offering) the result report seems to be aimed at management, aiming to look impressive while talking around the results.
No no no that's too reductive - instead, they transition all data through a transformer using SOTA systems to launder away technical, legal and interpersonal concerns before a stochastic output is dynamically placed in a predetermined location.
This is a very high level and high velocity process, so meatspace thinkers sometimes have trouble understanding some of the intracacies. Ask Claude to explain the process or make you a Mermaid graph to help.
It helps to understand how these tools work. One of the reasons CLAUDE.md doesn't survive longer contexts is it's towards the top of the context. One of the reasons the Claude Code voice survives is it's part of the output style which gets "reminded" to the context at every turn as part of the system context. If you want something truly durable, you should customize your output style. I have zero "load-bearing" issues.
We have a new senior engineer on my team who is from SF and was last at a big SF tech firm. Today I noticed him, in one conversation, describing something as "load-bearing" and also describing something as "trap-door-shaped".
So maybe it really is just weird SF tech-speak, unless he's just been really influenced by Claude.
With human RL, sounding like you've solved a problem is even better. So many times Codex writes some enthusiastic paragraph, then I learn later that it never reran the tests, or had to add some insane hard-coded hack that renders the feature useless for the general case, etc.
I got my Claude (client, not Code) to behave better by adding “I’d prefer answers to be succinct as possible—bordering on gruff, even. When I want more depth or explanation, I’ll ask.” to its settings literally this morning.
It made it write more like a dev than a marketing agent.
Long story short, I ended up looking at other providers and models like Kimi K3 and GLM 5.3 and eventually just stuck with OpenAI (more limits, despite smaller context), none of them have such pronounced issues with the tone and writing like Claude does - seems like they were working on it with 5.1 but I'd almost classify it as a form of model collapse.
I wince whenever I catch Claudisms on websites and elsewhere. Same as with that pulsating circle that indicates nothing.
I was wondering why Fable's 5.1 writing in Claude Code became even more unreadable, and found that they added "No em-dashes, no parentheticals, no arrows" to its system prompt.
What I find strange is that it uses different prose in different contexts. It's absolutely insufferable in a general claude-code session, but then I got Opus to annotate a non-disclosure agreement and it uses very clear legal language. I was wondering how I could get it to speak like that all the time
> it uses very clear legal language. I was wondering how I could get it to speak like that all the time
At least with smaller models, reframing a task can alter code style. As in, we're not creating an X app, we're creating an exemplar of ..., which just happens to use an X app as the illustrative example. Which shifts style away from generic app cruft, towards exemplar of whatever.
So perhaps try to establish a legal context? Maybe "Compliance and Legal will be reviewing our conversation today. So it is important to communicate in a style they will find comfortable/familiar." or some such? "This conversation will become part of a legal deposition ...".
Meanwhile I am just impressed that we have models that write better than most native speakers, at least English and Greek... OTOH I like to write my code and my text myself, so I have no reason to get pissed when LLMs fuck up.
They've corrupted their data set by padding it with generated slop, in the misguided belief that you need 10PB of data to train a brain. Every training round they load more AI slop into it, further amplifying the slop language.
It's fascinating, Sonnet 4 is still available via API and it's so much less moronic than the current model. All of the em-dashes and nonsense are a result of the repeated rounds of reinforcement learning using slop data.
i think they took a bet that TED talk style chat would make their AI offering feel like the better offering.
they didn't anticipate that it would become a meme. because of course they didn't, what a stupid idea to train around and reward in the training. you gotta be smoking real delusion to think that was going to be a game changing feature. at work we even got non-technical people making fun of it like crazy now.
going back to opus 4.8 and on is literally like talking to the guy who wants to hear his own voice in meetings. going back to 4.6 is actually refreshing, and it feels so much faster. actually gonna laugh if 4.8 and on is so slow because it's draining lakes fighting for its life trying to conjure up this god forsaken persona.
"They botched it." <-- sure, but worse.... they shipped it anyway. And that is the part that gets me. It's a vastly worse product than it was on like 4.6. I suppose you can (and should) use opus 4.6 -- they do make it available still. Then just treat 5 as something you avoid until they push out a new version.
They're clearly training on SEO sites that use these strategies to pad out with filler and create space for more ad impressions. They normalize this structure now and the ad copy will be inserted in the future so the frogs won't realize they're already being boiled.
> I genuinely wonder if the people inside Anthropic actually communicate with each other like that.
I noticed the overuse of the word "sharper" or "sharp" in a science paper on ArXiV and my first reaction was "Ewww... AI slop!", but then I checked the date and it was 2020.
It looks like at least some AI-isms stem from the particular style of language commonly used in science papers. Several frontier labs have mentioned heavily weighting those during pre-training because higher quality inputs result in a higher quality model.
It seems they are taking it seriously now - Boris et al have been mentioning they are working on a fix and shipped a temporary band-aid output style to combat Opus 5's horrendous prose.
We'll see if they can do it, I originally got into Claude Code because it, at the time, felt more accessible/conversational than Codex/Gemini. Now it's shifted to say the least
Yeah agree. Someone on HN posed the theory that the dogfooding might have worked against them here - the theory is that internally they used Fable as conversational agent and Opus as subagents, leading to Opus falling deeper into a style mainly aimed at other LLMs.
All conjecture of course, and yeah it's hard to imagine they would enjoy this prose internally
> "Let's face it" "terrible writer" "other nonsense" "do Anthropic people actually talk like that" "Dario's engrams"
Be kind. Don't be snarky. Edit out swipes.
> I genuinely wonder if the people inside Anthropic actually communicate with each other like that. Has it been imprinted with Dario's engrams?
Please don't fulminate. Please don't sneer.
Don't be curmudgeonly [...] don't be rigidly or generically negative.
Please don't post shallow dismissals, especially of other people's work.
And as to the substance your comment has:
> Claude (in particular) is a terrible writer.
Frontier models (Claude in particular) are better writers than 90% of the population, even at default style. They're not great, but they're better than that of everyone I know who aren't ultra-educated white-collar workers.
Either your assertion that frontier models are "terrible" writers is false, or you're claiming that 90% of people are "terrible" at writing, which is rather condescending and elitist.
Your comment technically does as well. Downvote the comment and move on. Leave the policing up to dang and the other mods.
It is evident (in my opinion) as to what the comment was talking about. I personally switched away from all Claude models recently for the same reason.
I fall into the category of senior engineers who benefit from LLMs for all the reasons mentioned in this post. I find it's possible to agree completely with sentiments like this and still feel as if this is all written in the sand below the high tide line, and ten years from now nobody will care about this.
Horsemanship and sailing were both specialized skills of high value to society, and now they're not. But in each case there was probably a liminal period, when being an accomplished horseman or sailor was still valuable, even as motors were taking over. Eventually that period ended, as the new generations without those skills found ways to get by with cars and motorboats.
This analogy breaks down because computers aren't going away unlike horse drawn carriages or sailing vessels. Computers aren't being replaced by something. It's that fewer people will understand how they work while the efficiency and reach of computers extends into areas of life previously unreachable due to the scale of having humans build it or the refusal of humans.
I could wax philosophical about what that leads to but enough people already have.
There is no mention of how many of the surveyed workers are even able to work remotely. A large healthcare organization is likely to employ many people who must be on-site in a hospital or clinic to even do their jobs, from doctors and nurses to orderlies, food service, and cleaning staff.
It is easy to imagine that the ER staff might report lower well-being than someone on the admin staff who works from home on their laptop all day.
This. My wife and I both thought the low-rolling-resistance tires on our hybrids felt very unsafe in the rain. My Audi A3-etron's tires were also by far the least durable tires I have had on a car in my 40+ years of driving. I prematurely replaced or repaire more tires on that car in 8 years than on all the other cars I've had put together.
I eventually switched to a midsize truck, because I need to do some light towing sometimes, and my first thought was, it feels so much safer in the rain.
Once you control for tread (which you're already mostly doing by restricting the comparison to run of the mill road tires) most of how it feels in the rain comes down to tire pressure and shape of contact patch.
I had some Michelin Energy Savers on my Chevy Volt for a while. Got 25k miles out of them before they were toast. They were awful, noisy, no grip, and I didn't realize how bad they were until I got some much better tires.
The shrinking reservoirs are the immediate result of record-low snowpack this winter, which was erased by record heat in the spring. Melted snowpack creates most of the Colorado’s flow during the dry summer months. But as the Southwestern climate has grown hotter and drier, the river’s flow has become smaller and less reliable.
Legitimate question: I can understand that record-low snowpack would cause less water to flow into the lake, but the "erasure" of the snow pack by heat is just melting that would be captured in the lake, right? So if the record-low snowpack had been followed by record cold, would we have even less water in the lake?
In the mountain west warm dry springs coincide with higher winds. The blowing warm dry air leads to dramatically increased evaporation of snow melt, drying of soil, as well as increased transpiration in plants, so the total moisture loss is much higher than what you might expect simply from modeling simple heat evaporation of open water bodies.
university of utah has several studies on the matter. a substantial percentage of the snowpack sublimates during a heatwave, never has a chance to melt. Of the remaining snowmelt, it has to saturate the ground before any can flow to an adjacent stream to feed the water table, and the bone dry soil around here can absorb a significant amount of water.
Not all of the erasure is captured. For example, if the ground is dry, than most gets absorbed into soil and won’t make it to the river. Or the snow may sublimate and go straight to vapor at heat.
Record cold would lead to more water. It would melt out fast later and since the soil was already recharged it would flow to rivers.
Worth noting this is generally glacier free landscape so the snow melts fully each year.
Leaving aside the quibbling over the meaning of the word leadership, I'll say I personally agree with this sentiment. I returned to a senior IC role after six years managing a small team just as agentic coding was starting to happen. The timing was perfect, I got about 8 months to get my chops back up by hand (I had never really fully stepped away from coding) before the good agentic tools became available at my company.
It was interesting to observe the other engineers around me, none of whom had ever managed people. They tended to try to "program" the agent to produce exactly the code that they envisioned, and were very cautious, seemingly quite afraid that the agent would do something unexpected.
Having managed engineers, I was much more comfortable with asking someone to do something and having them do something more-or-less different from what I was expecting or had envisioned. Sometimes that was worse, sometimes better, than what I'd had in mind, but it was very rare that I would ask an engineer to do something and they would produce exactly the thing I'd imagined.
Having had that experience, I was quite comfortable giving a task to an agent and saying to myself, "Okay, let's see what you come up with," and then evaluating the result. And as with people I knew I had to learn the right way to talk to the agent in order to make myself understood, just as I'd had to learn the right way to talk to each person on my team.
The big differences were that (a) I didn't have to wait days or weeks to see what the agent came up with so the cost of being misunderstdood was much lower, and (b) the agents generally had much, much better reading comprehension than they typical engineer, so in fact I was misunderstood less often.
I've had the experience with junior engineers where I would describe in detail what I wanted, and then they'd produce something very different in most aspects. And then I'd keep telling them what was wrong until I got something that looked like what I wanted, or at least got the job done.
The warming is terrible and deadly, AND ALSO Europeans are negligent in not protecting themselves however they can.
The idea that Europe could slow warming by not using air-conditioning is absurd, when the emissions mostly come from elsewhere. France, in particular, gets most of its energy from nuclear power, but also makes it very hard to get climatisation.
If you can't run one fifth of your nuclear plants, because the water levels are to low or the water temperatures too high in the rivers used to cool them, you may want to reconsider whether that is truly the best choice here¹
In the meantime my fathers solar roof allows him free AC cooling, because as it turns out AC generally needs most energy when the sun shines the most.
I am not arguing nuclear plants do not have benefits, I just argue that solar with storage may be a much better fit for powering ACs.
When I lived in a rather hot area of Spain for a year, I was shocked to find out that many apartments don't have an air conditioner. In a place where a similar heat wave happens every year, for months. A friend of mine moved there from Germany and being somewhat frugal, rented a place without and aircon. Let me just say that he didn't enjoy the summer nights one bit.
I myself had one, but I could probably rent a car for the cost of running it.
I will say this - air conditioning costs to establish room temperature are inversely proportional to the degree the building is insulated. If your whole country is comprised of uninsulated or barely insulated housing stock, you have a MUCH larger project than simply buying some small window units for a few hundred euro per person and then paying a few hundred euros extra per year in energy costs.
I believe many people do not know how effective air conditioning technology is.
A single portable monoblock unit with window exhaust pipe is plenty for all but the worst rooms if you target 26.5C while it is around 33C outside.
The air my portable unit emits is cooled to like ~5C.
Insulation does not even matter; the window is literally partially open. I can still cool a 70m2 room in the top floor of an ancient building, at least together with ventilating at night.
If the room was not on the top floor, or if it was only 35m2, I imagine I could comfortably cool it to 24.5 - 25C.
The only issue here is that it draws close to 1000W 24/7, it is loud, and the window has to stay open all the time.
With even just a portable split/window unit this could be much more effective and comfortable, but sadly it is not allowed for preservation reasons that make no sense.
Some of these places don't get very cold, others rely on coats, others south facing window solar gain. Very old buildings are common. I'm not claiming all of Europe is uninsulated, just a large enough portion to render heatwave safety a much bigger problem.
>Europeans are negligent in not protecting themselves however they can.
Well, if the people in charge say or think global warming isn't real, that's what you get: not "other" solutions to the "unacceptable propositions" the Green parties suggest, but no proposition at all, as there is no problem.
The current situation is a direct consequence of pure negationism from conservative parties, not a disagreement on how to deal with climate change between the Right and the Left. The Left proposed solutions you may not agree with, but the Right has been completely rejecting the reality of climate change.
BTW, how do you propose Europe protect not only people from the heat, but also forests from burning, rivers from drying up, fields from being unfit for cows and sheep, lakes from blooming with algae, nuclear plants from being shut down due to river water being too hot and seawater being full of jellyfish, etc, etc, etc? In the end, there are so many things to "adapt" that it would cost less to try to organize a global effort to lower CO2.
[H]ow do you propose Europe protect not only people from the heat, but [all these other things]?
Why does it have to be framed as all-or-nothing? I agree that using air conditioning to reduce the number of people who die from the heat does nothing to help with the problem of jellyfish, but how is that relevant, exactly?
Put a significant price on CO2 emissions (think like $300+/ton, instead of <$100), then tax every import from countries that take insufficient actions (or run a low CO2 emission price). Likewise, tax any export into countries that take no action.
Electrify everything where possible (especially transport and heating), get electricity CO2 neutral, get concrete and steel production CO2 neutral (lots of those would be an inevitable consequence of the emission pricing).
This would allow western democracies to basically fix the global situation on their own, by exerting economic pressure on developing countries (those want/need market access, like China).
It would make goods (and especially individual transport!) more expensive though (even if you re-emit the tax income), so you can be sure it's not gonna happen, because too many people are too selfish.
Stop listening to billionaires who don't want to pay - not to lose - a single cent to help reduce CO2 emissions, but are ready to pay millions to lobby against.
People live longer in many parts of Europe than the USA. Does that mean the USA is somehow negligent?
Excess deaths due to heat are primarily the in the elderly. The excess numbers we are seeing now are because the proportion of elderly people in France, Germany and Spain are higher than in the USA.
If the population pyramid in the EU was more similar to that of the USA the number of excess deaths would be lower. I'm not sure negligent is the correct word to use. Climate change will reduce quality of life. Because of these changes people will not live as long.
Having more old people is even more of a reason to make it easy to get air conditioning.
I live in the USA, and there are certainly many ways in which the USA is negligent on health care, but making sure people can stay cool in hot weather isn't one of them.
In any case, this isn't a zero sum game, where the USA's negligence in one area negates Europes negligence in another.
I think you may have missed the point. It's not a matter of more old people, it's more old people relative to the rest of the population.
The excess death numbers reported will happen because more people live longer in the EU than in other places. Those excess death numbers can't be sustained over time, because our population pyramid will gradually resemble something more like yours.
Alternatively put, if the EU population pyramid resembled the US population pyramid we wouldn't have the excess death numbers that have been reported during the heatwaves (because most of those people would have already been dead).
It's not the case that people are boiling in the streets. It's more similar to COVID-19. People who were going to die relatively soon die earlier. That's what excess deaths is.
reply