leemoore

if you are refreshing a cache of 120k tokens, you have to input the same 120k tokens. you can't bounce the cache with 1 byte. A proper cache hit requires your entire context that is cached. if you change 1 byte in it, everything after that byte is a cache write and no longer a cache hit

it could be either. subscription defaults to 60 min with 2x cache writes. extra usage and api defaults to 5 min with 1.25 cache writes. You can override either with settings

If you're running the subscription, by default you are paying 2x for cache writes and you're getting an hour for expiry. So refreshing based on 5 min is wasteful. You need to detect whether you are in 5 min or 1 hour mode.

The post would have been stronger and more useful focusing and expanding on the abstraction portion of his message as well as system and program design. Those elements were there but seem sidelined by the focus on DSL. For most developers on most projects, the key take away is we all have to both get good at system and program design especially around abstractions, shapes, responsilibility. Then we have to get good at inserting our selves in the AI loop to steer those decisions a lot in the beginning, regularly as the project kicks off and still more than we think as the project matures.

The counter intuitive pattern I see emerging is if you can cleanly determine intent, of the call you fix the call and prepend informative text to the tool call response indicating the mistake made and how to fix in the future then followed by the actual tool call. In this case you can validate fields and rather than throw a hard error determine if it's an extra field that isn't needed. If so you correct the call and prepend a corrective response in the tool call. This saves turns, it instructs the model in context so less likely to happen later and helps models that aren't so good at recovering from bad tool calls and staying on their longer horizon agentic task (most non openai and anthropic models)

GLM 5.2 feels like Opus 4.6 level. I actually think 4.6 and GLM work better in practice than opus 4.7 or 4.8 as I find both of those more erratic and seem to randomly have a super dumb turn. That random bad turn I see doesn't seem to be hitting the benchmark scores but they make 4.7 and 4.8 very hard to use for me. GLM is more stable like opus 4.6

For developers at non tech fortune 500 companies, I would put money on Windows being the primary workstation os by a lot

It's the executive branch asserting control in this space and requiring all SOTA model providers to bend the knee. Anthropic is the least capable of playing the bend the knee game so is getting the first and worst smack down

If you don't have the capacity to have your mind changed through friction and disagreement with a SOTA LLM and feel compelled to frame those who do to through absurdly reductive statement like "insane arguing with a machine" then that says more about your limitation and lack of understanding than the OP's or Claudes.

I have the same processor and ram. The dense 30b ish Gemma/Qwen really don't break 10 TPS with or without MTP. MOE's in this range feel more usable if they are smart enough for your work. Probably would still use hosted versions of these over local unless. MOE's feel somewhere between sonnet 3.5 and 3.7 to me. Dense feels between sonnet 3.7 and 4 in basic coding or local agentic capabilities (not close to those in chat or world knowledge)

The way you do this (and the way opencode does it) is you do most of your pruning in more recent history. Last I looked at opencode, they start pruning tool call results after 2 full agentic turns. So you probably dont get quite as good hits on cache for the most recent 1-5% of your turns, but after that everything else caches fine and those tool calls that likely aren't relavent to your session anymore are gone.

Enterprise customers aren't running 20 bucks a month for claude pro subscriptions. My company provides developers about 1k worth of usage limits a month and best I can tell they get maybe a 30% savings off of API cost tops. That's not an insane subsidy. Many other jobs titles are only allowed 50 a month and those folks are constantly running out.

Github Copilot has been doing this with business and enterprise seats, but that will be coming to a head very soon. I expect a fast follow after june when they re-align consumer pro and pro+ accounts.

OpenAi seems to be trying to throw tokens at clients to get lock in. So i'd be most worried about the rug pull that will come from open AI post IPO. Anthropic is already acting responsibly in this area and github copilot is attempting to remediate their insane subsidies in the next several months.

Gemini is bad at this sort of thing but I find all models tend to do this to some degree. You have to know this could be coming and give it indicators to assume that it’s training data is going to be out of date. And it must web search the latest as of today or this month. They aren’t taught to ask themselves “is my understanding of this topic based on info that is likely out of date” but understand after the fact. I usually just get annoyed and low key condescend to it for assuming its old ass training data is sufficient grounding for correcting me.

That epistemic calibration is is something they are capable of thinking through if you point it out. But they aren’t trained to stop and ask/check themselves on how confident do they have a right to be. This is a meta cognitive interrupt that is socialized into girls between 6 and 9 and is socialized into boys between 11-13. While meta cognitive interrupt to calibrate to appropriate confidence levels of knowledge is a cognitive skill that models aren’t taught and humans learn socially by pissing off other humans. It’s why we get pissed off st models when they correct ua with old bad data. Our anger is the training tool to stop doing that. Just that they can’t take in that training signal at inference time

What demographic are you in that is leaving anthropic in mass that they care about retaining? From what I see Anthropic is targeting enterprise and coding.

Claude Code just caught up to cursor (no 2) in revenue and based on trajectories is about to pass GitHub copilot (number 1) in a few more months. They just locked down Deloitte with 350k seats of Claude Enterprise.

In my fortune 100 financial company they just finished crushing open ai in a broad enterprise wide evaluation. Google Gemini was never in the mix, never on the table and still isn’t. Every one of our engineers has 1k a month allocated in Claude tokens for Claude enterprise and Claude code.

There is 1 leader with enterprise. There is one leader with developers. And google has nothing to make a dent. Not Gemini 3, not Gemini cli, not anti gravity, not Gemini. There is no Code Red for Anthropic. They have clear target markets and nothing from google threatens those.

Gemini isn't code red for Anthropic. Gemini threatens none of Anthropic's positioning in the market.

Not sure about parent, but my current bar is set by GPT-5 high in codex cli. Sonnet 4.5 doesn't quite get there in many of the use cases that are important to me. I still use sonnet for most less intelligence phases and tasks (until I get crunched by rate limits). But when it comes to writing the final coding prompt and the final verification prompt and executing a coder or a verifier that will execute and verify well it's GPT 5 high all the way. Even if sonnet is better at tool calling, GPT 5 High is just smarter and has better coding/engineering judgement and that difference is important to me. So I very much get the sentiment of not going below sonnet intelligence 4.5 for coding. It's where I draw the line too.

My success and experience generally matches yours (and the authors'). Based on my experience over the last 6 months, nothing here around more senior developers getting more productivity and why is remotely controversial.

It's fascinating how a report like yours or theirs acts as a lightning rod for those who either haven't been able to work it out or have rigid mental models about how AI doesn't work and want to disprove the experience of those who choose to share their success.

A couple of points I'd add to these observations: Even if AI didn't speed anything up... even if it slowed me down by 20%, what I find is that the mental load of coding is reduced in a way that allows me to code for far more hours in a day. I can multitask, attend meetings, get 15 minutes to work on a coding task, and push it forward with minimal coding context reload tax.

Just the ability to context switch in and out of coding, combined with the reduced cognitive effort, would still increase my productivity because it allows me to code productively for many more hours per week with less mental fatigue.

But on top of that, I also antectodally experience the 2-5x speedup depending on the project. Occasionally things get difficult and maybe I only get a 1.2-1.5x speedup. But it's far easier to slot many more coding hours into the week as an experienced tech lead. I'm leaning far more on skills that are fast, intuitive abilities built up from natural talent and decades of experience: system design, technical design, design review, code review, sequencing dependencies, parsing and organizing work. Get all these things to a high degree of correctness and the coding goes much smoother, AI or no AI. AI gets me through all of these faster, outputs clear curated (by me) artifacts, and does the coding faster.

What doesn't get discussed enough is that effective AI-assisted coding has a very high skill ceiling, and there are meta-skills that make you better from the jump: knowing what you want while also having cognitive flexibility to admit when you're wrong; having that thing you want generally be pretty close to solid/decent/workable/correct (some mixture of good judgement & wisdom); communicating well; understanding the cognitive capabilities of humans and human-like entities; understanding what kind of work this particular human/human-like entity can and should do; understanding how to sequence and break down work; having a feel for what's right and wrong in design and code; having an instinct for well-formed requirements and being able to articulate why when they aren't well-formed and what is needed to make them well-formed.

These are medium and soft skills that often build up in experienced tech leads and senior developers. This is why it seems that experienced tech leads and senior developers embracing this technology are coming out of the gate with the most productivity gains.

I see the same thing with young developers who have a talent for system design, good people-reading skills, and communication. Those with cognitive flexibility and the ability to be creative in design, planning and parsing of work. This isn't your average developer, but those with these skills have much more initial success with AI whether they are young or old.

And when you have real success with AI, you get quite excited to build on that success. Momentum builds up which starts building those learning skill hours.

Do you need all these meta-skills to be successful with AI? No, but if you don't have many of them, it will take much longer to build sufficient skill in AI coding for it to gain momentum—unless we find the right general process that folks who don't have a natural talent for it can use to be successful.

There's a lot going on here with folks who take to AI coding and folks who dont. But it's not terribly surprising that it's the senior devs and old tech leads who tend to take to it faster.

Same, if you dont give opus big enough problems it's more likely to go off the rails. Not much more likely but a little more likely

One challenge around studying these drugs is efficacy is often linked to expectation. Some speculate 1 key mechanism at work is an amplification of the placebo effect. If this is the case then clearly testing of these drugs using traditional techniques to eliminate the placebo effect will be problematic

This binary assertion is fairly broadly given and I think it’s a narrow view that is a subtle form of unintentional disinformation that endangers. Any reduction of viral load on exposure is by nature better. The intensity of initial viral Old can help determine whether you get covid 19 and how strong it hits. It’s difficult to believe that some worth of cloth covering will have absolutely zero reduction of viral load. No you shouldn’t think it’s a comprehensive protection. But it’s silly to handwave away as useless as many do. Anything that limits spread should be encouraged. People still need to minimize going out, but if they have to, and have nothing else, dismissing all other masks as useless is not helpful.

Stream Entry is a Theravada Buddhist term for the first of 4 stages or paths of awakening. A precise model has been created but it tends to be only useful if you have a certain mental makeup and you engage in the practices with a certain degree of intensity. Even then it feels to me like too much effort to make a subjective process set of experiences into a hard and fast model masquerading as authoritative and true.

I'm not so familiar with your stages model so I can't really comment. Based on your last paragraph, my engagement with meditation comes from a very different place than yours. It started in order to stabelize attention after an ADHD diagnosis; a way to get off meds. Quickly that was left behind and the real focus was enlightenment, spiritual awakening or whatever you want to call it. It was motivated by a deep spontaneous inner faith that there was something to it and it was important that I go 100% into it and discover for myself. It was not based on research or attempting to find the best approach to make my mind measurably more effective. I just felt a powerful internal yearning for awakening and I had to follow it.

I just popped over there. I haven't been in a while. It looks like one of my longer reports was linked up to their wiki. This chronicles my meditative exploits from when I really got into it up until an 8 month sabbatical I took and had some initial breakthroughs. http://www.dharmaoverground.org/web/guest/discussion/-/messa...

To take this phenomenon further down the Rabbit Hole, if you pay close enough attention, a defined sense of self is just a feeling or thought that rises from time to time. However addiction/compulsion to perpetual thinking (which 99.99% of humans are caught in) plus the smoothing programs that provide a sense of continuousness and continuity begin to act on that sense of self firmly establishing a deep continuous sense of me. And there is a tremendous amount of conscious energy wrapped up in the perpetuation of a continuous sense of me. Certain states and activities more deeply lock the sense of me in including anger, conflict, proving yourself right, proving someone wrong.

Meditation and Hallucinogens can quiet the mind (or more dramatically short circuit habitual mind loops) and allow you to see the gaps in the self-ing process. The self-ing process never really stops. You never really stop having a sense of self arise from time to time, but when you see the frames and gaps of self-ing you begin to question a lot of who you think you are. And you begin to see the amount of time and energy you spend protecting the self from paper tigers.

Spiritual circles generally call this awakening and tend to over-emphasize it's specialness and idealize those who have deeply conditioned themselves to reside in that state. But it is real and mostly desirable phenominon (with a few drawbacks). Putting aside questions of full enlightenment and mind blowing transcendent unitive states, simply freeing up the consciousness that gets locked up in perpetuating the continuous sense of self and all the efforts to defend as if you are defending your physical body from harm allows for a richer, higher resolution and more textured moment to moment experience of life.

From my subjective experience it is the same sorts of filters that smooth out reality, that also give you a continuous sense of a psychological self which basically manufactures the vast majority of our problems and suffering.

I don't have that. However if your not familiar with the Dharma Overground, I'd check it out. There are lots of dedicated geeky practitioners over there that share experiences and to whatever degree possible attempt to document the results with more scientific rigor than a typical spiritual community. http://www.dharmaoverground.org/

I've had something similar since at least my teens. At night if I look closely at streetlights, I see what looks to me like a magnetic field around streetlights. I don't think that's what I'm actually seeing, but that's what it looks like.

I believe the main factor at play in the first pair of images, is that on LSD and other hallucinogens as well as intensive meditation is that there are little mind programs that run on our perception that smooth out reality for us.

If you notice, even just sitting still we are constantly moving our head just a little, constantly moving our eyes. Without these programs, our visual field would seem far more jumpy and unstable. In certain states of consciousness, these smoothing programs can become intermittent, less effective or even completely disabled.

Also, when mind rambling, chatter and perpetual loops stop, a greater amount detail can be seen and perceived through all senses. At higher and higher levels of perceptual resolution with our filters disabled and the smoothing programs down, you begin to notice that we don't perceive reality smoothly. We perceive it in tiny little frames. If you just watch the first picture, your mind isn't drawn to he frames. After looking for a bit at the second picture of the pair, suddenly you can start to see the frames of perception more clearly. It's simply a matter of learning to (or being tricked to) get past our habitual programs and filters to tune into a more fine grained perceptual reality happening.

These observations come from my own experience with psychedelics in my 20's and extensive meditation practice including a number of longer meditation retreats.

Theravada Buddhist Practitioner here. In addition to daily meditation practice, I've attended 8 months worth of meditation retreats in the last 4 years. These retreats consisted of 8-15 hours a day of formal meditation (depending on how hard I was going at it) with an effort to continuity of mindfulness in between formal sessions. At this point, I estimate I have about 3-4k of my 10k hours required for high mastery.

My understanding of the purpose and benefit of practice is to observe experience as it arises (whether physical sensations, sensory input or thoughts) as it arises and see how you react to it.

One key insight you find is we tend to contract toward things that are pleasant and contract away from experience that is unpleasant and space out (or seek distraction) during experience that is neutral or boring. A cognitive understanding of this contraction and spacing out is insufficient to decondition the human brain from doing it. The conditioning is too deep.

It requires using concentration and mindfulness as a microscope on our perception to see and feel the additional pain we create in the contraction. As we repeatedly connect with the pain of contraction we slowly over time stop doing it and begin to experience life in a whole new way. The capacity to face pleasant, unpleasant and neutral experiences without contracting or spacing out is quite literally a life changer.

A notion of pure meditation doesn't have much meaning for me. It just sounds like someones attempt to take an extremely broad and varied set of useful practices and overly simplify in an effort to assert internally or externally that there is one right/best/pure/perfect way of doing something. I don't subscribe to that view. Like programming languages or frameworks or methodologies or any other set of maps and techniques they have their various levels of usefulness depending on person and circumstance.

Also, the idea of giving up desire, passion and things of this nature is often misunderstood. It is not necessary to give up ones humanity, emotions, loves or anything else. It's more about changing the way in which you perceive and engage with all experiences of life. It's not like I've stopped feeling bad. I feel bad plenty. It's more that feeling bad (and feeling good) is experienced in a very different way.

I've found that intensive practice of meditation over the last 3 years including both a daily practice and going on more extended retreats has fundamentally shifted the way in which I perceive and engage with moment to moment experience which has in turn steadily raised my baseline level of happiness.

I fully understand this is subjective anecdotal data and that it's quite possible I'm fooling myself though I don't believe I am. I've heard similar reports from others who have engaged deeply with various contemplative practices.