Then hand over to Claude Sonnet.

With hard requirements listed, I found out that the generated code missed requirements, had duplicate code or even unnecessary code wrangling data (mapping objects into new objects of narrower types when won't be needed) along with tests that fake and work around to pass.

So turns out that I'm not writing code but I'm reading lots of code.

The fact that I know first hand prior to Gen AI is that writing code is way easier. It is reading the code, understanding it and making a mental model that's way more labour intensive.

Therefore I need more time and effort with Gen AI than I needed before because I need to read a lot of code, understand it and ensure it adheres to what mental model I have.

Hence Gen AI at this price point which Anthropic offers is a net negative for me because I am not vibe coding, I'm building real software that real humans depend upon and my users deserve better attention and focus from me hence I'll be cancelling my subscription shortly.

The market-leading technology is pretty close to "good enough" for how I'm using it. I look forward to the day when LLM-assisted coding is commoditized. I could really go for an open source model based on properly licensed code.

but then two months ago 4.6 started getting forgetful and making very dumb decisions and so on. Everyone started comparing notes and realising it wasn’t “just them”. And 4.7 isn’t much better and the last few weeks we keep having to battle the auto level of effort downgrade and so on. So much friction as you think “that was dumb” and have to go check the settings again and see there has been some silent downgrade.

We all miss the early days of 4.6, which just show you can have a good useful model. LLMs can be really powerful but in delivering it to the mass market Anthropic throttle and downgrade it to not useful.

My thinking is that soon deepseek reaches the more-than-good-enough 4.6+ level and everyone can get off the Claude pay-more-for-less trajectory. We don’t need much more than we’ve already had a glimpse of and now know is possible. We just need it in our control and provisioned not metered so we can depend upon it.

API Error: Claude's response exceeded the 32000 output token maximum. To configure this behavior, set the CLAUDE_CODE_MAX_OUTPUT_TOKENS environment variable.

For reasons that continue to elude me, almost exactly one year ago, Anthropic cancelled my Claude Pro plan. To appeal, you must fill out a Google docs form. And wait. In my case, I’ve waited for about one year. Once I managed to email with a human but they quickly plugged that hole with a chatbot that sends you back to their never-to-be-reviewed form. No route to escalate.

A year gives one a long time to think about things. Maybe it was because I was on a VPN temporarily. Otherwise, no clue. I’m a hobbyist embedded developer. That’s it.

So no, Anthropic support isn’t just poor; it’s nonexistent.

GPT 5.4+ takes its time and considers even edgecases unprovoked that in fact are correct and saves me subsequent error hunting turns and finally delivers. Plus no "this doesn't look like malware" or "actually wait" thinking loops for minutes over a oneliner script change.

Now I'm looking for an extremely simple open-source coding agent. Nanocoder doesn't seem install on my Mac and it brings node-modules bloat, so no. Opencode seems not quite open-source. For now, I'm doing the work of coding agent and using llama_cpp web UI. Chugging it along fine.

> “you can’t be serious — is this how you fix things? just WORKAROUNDS????”

If this is how you’re interacting with your agents I think you’re in for a world of disappointment. An important part of working with agents is providing specific feedback. And beyond that making sure this feedback actually available to them in their context when relevant.

I will ask them why they made a decision and review alternatives with them. These learnings will aid both you and the agent in the future.

I haven't seen anyone mention this publicly, but I've noticed that the same model will give wildly different results depending on the quantization. 4-bit is not the same as 8-bit and so on in compute requirements and output quality. https://newsletter.maartengrootendorst.com/p/a-visual-guide-...

I'm aware that frontier models don't work in the same way, but I've often wondered if there's a fidelity dial somewhere that's being used to change the amount of memory / resources each model takes during peak hours v. off hours. Does anyone know if that's the case?

I tried Kimi 2.6 and it's almost comparable to Opus. Anthropic lost the ball. I hope this is a sign the we are moving towards a future where model usage is a commodity with heavy competition on price/performance

One group is consistently trying to play whack-a-mole with different models/tools and prompt engineering and has shown a sine-wave of success.

The other group, seemingly made up of architects and Domain-Driven Design adherents has had a straight-line of high productivity and generating clean code, regardless of model and tooling.

I have consistently advised all GenAI developers to align with that second group, but it’s clear many developers insist on the whack-a-mole mentality.

I have even wrapped my advice in https://devarch.ai/ which has codified how I extract a high level of quality code and an ability to manage a complex application.

Anthropic has done some goofy things recently, but they cleaned it up because we all reported issues immediately. I think it’s in their best interests to keep developers happy.

My two cents.

First was the CC adaptive thinking change, then 4.7. Even with `/effort max` and keeping under 20% of 1M context, the quality degradation is obvious.

I don't understand their strategy here.

Here is a sample report that tries out the cheaper models + the newest Kimi2.6 model against the 5.4 'gold' testcases from the repo: https://repogauge.org/sample_report.

I use AI, but only what is free-of-charge, and if that doesn't cut it, I just do it like in the good old times, by using my own brain.

I am certainly not saying people should “spend more money,” more like the Claude Code access in the Pro plan seems kind of like false advertising. Since it’s technically usable, but not really.

My experience very suddenly and very clearly degraded over the last few days.

Today I was trying to build a simple chess game. Previous one shots were HTML, this gave me a jsx file. I asked it to put it HTML and it absolutely devoured my credits doing so, I had to abort and do it manually. The resulting app didn't work, and it had decided that multiplayer could work by storing the game state only on local storage without the clients communicating at all

And by crikey do I empathise with the poor support in this article. Nothing has soured me on Anthropic more than their attitude.

Great AI engineers. Questionable command line engineers (but highly successful.) Downright awful to their customers.

There's really no immediate solution to this other than letting the price float or limiting users as capacity is built out this gets better.

All mostly mitigatable by rigorous audits and steering, but man, it should not have to be.

The 20$ plan has incredible value but also, the limit are just way too tight.

I'm glad Claude made me discover the strength of ai, but now it's time to poke around and see what is more customer friendly. I found deepseek V4 to be extremely cheap and also just as good.

Plus I get the benefit to use it in vs code instead of using Claude proprietary app.

I think that when people goes over the hype and social pressure, anthropic will lose quite a lot of customer.

The first job of any support system—both in terms of importance and chronologically—is triage. This is not a research issue and it's not an interaction issue. It's at root a classification problem and should be trained and implemented as such.

There are three broad categories of interaction: cranks, grandmas, and wtfs.

Cranks are the people opening a support chat to tell you they have vital missing information about the Kennedy Assassintion or they want your help suing the government for their exposure to Agent Orange when they were stationed at Minot. "Unfortunately I can't help with that. We are a website that sells wholesale frozen lemonade. Good luck!"

Grandma questions are the people who can't navigate your website. (This isn't meant to be derogatory, just vivid; I have grandma questions often enough myself.) They need to be pointed toward some resource: a help page, a kb article, a settings page, whatever. These are good tasks for a human or LLM agent with a script or guideline and excellent knowledge/training on the support knowledge base.

WTFs are everything else. Every weird undocumented behavior, every emergent circumstance, every invalid state, etc. These are your best customers and they should be escalated to a real human, preferably a smart one, as soon as realistically possible. They're your best customers because (a) they are investing time into fixing something that actually went wrong; (b) they will walk you through it in greater detail than a bug report, live, and help you figure it out; and (c) they are invested, which means you have an opportunity for real loyalty and word-of-mouth gains.

What most AI systems (whether LLMs or scripts) do wrong is that they treat WTFs like they're grandmas. They're spending significant money on building these systems just to destroy the value they get from the most intelligent and passionate people in their customer base doing in-depth production QC/QA.

I tried Claude recently and it was able to one-shot fixes on 9/9 of the bugs I gave it on my large and older Unity C# project. Only 2/9 needed minor tweaks for personal style (functionally the same).

Maybe it helps that I separately have a CLI with very extensive unit tests. Or that I just signed up. Or that I use Claude late in the evenings (off hours). I also give it very targeted instructions and if it's taking longer than a couple minutes - I abort and try a different or more precise prompt. Maybe the backend recognizes that I use it sparingly and I get better service.

The author describes what sounds like very large tasks that I'd never hand off to an AI to run wild in 2026.

Anyway I thought I'd give a different perspective than this thread.

For actual code that goes out to production, I generally figure out how I would solve the problem myself (but will use Claude to bounce ideas and approaches -- or as a search engine) and then have Claude do the boring bits.

Recently I had to migrate a rules-engine into an FSM-based engine. I already had my plan and approach. I had Claude do the boring bits while I implemented the engine myself. I find that Claude does best when you give it small, focused, incremental tasks.

I occasionally ask AI to write lots of code such as a whole feature (>= medium shirt size) or sometimes even bigger components of said feature and I often just revert what it generated. It's not good for all the reasons mentioned.

Other times I accept its output as a rough draft and then tell it how to refactor its code from mid to senior level.

I'm sure it will get better but this is my trust level with it. It saves me time within these confines.

Edit: it is a valuable code reviewer for me, especially as a solo stealth startup.

For work, unlimited usage via Bedrock.

Yes I’d like to get more usage out of my personal sub, but at 20/mo no complains

I think even with the worse limits people still hated it but when you start to either on purpose or inadvertently make the model dumber that's when there's really no purpose to keep using Claude anymore.

Even a simple prompt focused on two files I told Claude to do a thing to file A and not change file B (we were using it as a reference).

Claude’s plan was to not touch file B.

First thing it did was alter file B. Astonishing simple task and total failure.

It was all of one prompt, simple task, it failed outright.

I also had it declare that some function did not have a default value and then explain what the fun does and how it defaults to a specific value….

Fundamentally absurd failures that have seriously impacted my level of trust with Claude.

The thing is running local LLMs will give some kind of reliability and fixed expectations that saves a lot of time - yeah sure Claude might be fantastic one day, but what do I do when the same workload churns out shit the next and I am halfway thru updating and referencing a 500 document set?

Better the devil you know and all that.

Pro is gone. OpenAI plans are more expensive. He can only buy a Kimi plan, which is at least better than Sonnet. But frontier for cheap is gone. Even copilot business plans are getting very expensive soon, also switching to API usage only.

Before the fixes, they were complete trash and I was ready to cancel this month.

Now, I'm feeling like the AI wars are back -- GPT 5.5 and Opus 4.7 are both really good. I'm no longer feeling like we're using nerfed models (knock on wood)!

The filesystem tool cannot edit xml files with <name></name> elements in it

Most of this is about the billing system, which is apparently broken.

https://www.anthropic.com/engineering/april-23-postmortem

I was worried about Anthropic models quality varying and about Anthropic jacking up prices.

I don't think Claude Code is the best agent orchestrator and harness in existence but it's most widely supported by plugins and skills.

I'm debating trying out Codex, from some people I hear its "uncapped" from others I hear they reached limits in short spans of time.

There's also the really obnoxious "trust me bro" documentation update from OpenClaw where they claim Anthropic is allowing OpenClaw usage again, but no official statement?

Dear Anthropic:

I would love to build a custom harness that just uses my Claude Code subscription, I promise I wont leave it running 24/7, 365, can you please tell me how I can do this? I don't want to see some obscure tweet, make official blog posts or documentation pages to reflect policies.

Can I get whitelisted for "sane use" of my Claude Code subscription? I would love this. I am not dropping $2400 in credits for something I do for fun in my free time.

What I don't understand is these loud "voting with money" comments. What they are canceling is very subsidized plan to buy something that delivers a lot of value.

There are only two providers that can provide this level of models at very subsidized price - anthropic and openai. Both of them are bad in terms of reliability.

So I wonder what these people do after they "cancel" both of them? Do they see producing less result at same hourly rate as everyone else on the market as viable option?

Asked support hey i got nothing back i tried prompting several times used a ton of usage and it gave no response. I'd just like usage back. What I payed for I never got.

Just bot response we don't do refunds no exceptions. Even in the case they don't serve you what your plan should give you.

I’m blown away by how good it is lately

WTF are y'all doing that chews tokens so fast? I mean, sure, I could spin up Gas Town and Beads and produce infinite busy work for the agents, but that won't make useful software, because the models don't want anything. They don't know what to build without pretty constant guidance. Left to their own devices, they do busy work. The folks who "set and forget" on AI development are producing a whole lot of code to do nothing that needed doing. And, a lot of those folks are proud of their useless million lines of code.

I'm not trying to burn as many tokens as a possible, I'm trying to build good software. If you're paying attention to what you're building, there's so many points where a human is in the loop that it's unusual to run up against token limits.

Anyway, I assume that at some point they have to make enough money to pay the bills. Everything has been subsidized by investors for quite some time, and while the cost per token is going down with efficiency gains in the models/harnesses and with newer compute hardware tuned for these workloads, I think we're all still enjoying subsidized compute at the moment. I don't think Anthropic is making much profit on their plans, especially with folks who somehow run right at the edge of their token limit 24/7. And, I would guess OpenAI is running an even lossier balance sheet (they've raised more money and their prices are lower).

I dunno. I hear a lot of complaining about Claude, but it's been pretty much fine for me throughout 4.5, 4.6 and 4.7. It got Good Enough at 4.5, and it's never been less than Good Enough since. And, when I've tried alternatives, they usually proved to be not quite Good Enough for some reason, sometimes non-technical reasons (I won't use OpenAI, anymore, because I don't trust OpenAI, and Gemini is just not as good at coding as Claude).

I'm an executive, the devs complaining are getting retrained or put on the chopping block.

My rockstars are now random contractor devs from Vietnam. The aloof FTE grey beards saying "I don't know, it doesn't work very good on X." Are getting a talking to or being sidelined/canned. So far most of my grey beards are adapting pretty well.

I'm not waiting on people to write code any more. No way in hell.

It's almost unusable

We probably hit peak generative AI last year, now they probably use AI to improve the AI so it’s kinda garbage in garbage out, or maybe anthropic is deprioritizing users while favoring enterprise or even government where it provides better quality for higher contracts.

Oh wait, I don’t have to imagine. That’s what Anthropic does. A nice preview for what is in store for those who chose to turn off their brains and turn on their AI agents.

From "yay, claude is awesome" to "damn, it sucks". This is like with withdrawal symptoms now.

My approach is much easier: I'll stay the oldschool way, avoid AI and come up with other solutions. I am definitely slower, but I reason that the quality FOR other humans will be better.

AI used to be, the punched card replicator... its all replaceable.

I'm pretty sure it used to warn when you got close to your 5hr limit, but no, it happily billed extra usage. Granted only about $10 today, but over the span of like 45 minutes. Not super pleased.

Then within the last few months everything changed and went to shit. My trust was lost. Behavior became completely inconsistent.

During the height of Claude's mental retardation (now finally acknowledged by the creators) I had an incident where CC ran a query against an unpartitioned/massive BQ table that resulted in $5,000 in extra spend because it scanned a table which should have been daily partitioned 30 times. 27 TB per scan. I recall going over and over the setup and exhaustively refining confidence. After I realized this blunder, I referred to it in the same CC session, "jesus fucking christ, I flagged this issue earlier" -- it responded, "you did. you called out the string types and full table scans and I said "let's do it later." That was wrong. I should have prioritized it when you raised it". Now obviously this is MY fault. I fucked up here, because I am the operator, and the buck stops with me. But this incident really galvinized that the Claude I had come to vibe with so well over the last N months was entirely gone.

We all knew it was making making mistakes, becoming fully retarded. We all felt and flagged this. When Anthropic came out and said, "yeah ... you guys are using it wrong, its a skill issue" I knew this honeymoon was over. Then recently when they finally came out and ack'd more of the issues (while somehow still glossing over how bad they fucked up?) it was the final nail. I'm done spending $ on Anthropic ecosystem. I signed up for OpenAI pro $200/mo and will continue working on my own local inference in the meantime.

Anthropic can't even scale their own infrastructure operations, because it does not exist and they do not have the compute; even when they are losing tens of billions and can nerf models when they feel like it.

Once again, local models are the answer and Anthropic continues to get you addicted to their casino instead of running your own cheaper slot machine, which you save your money.

Every time you go to Anthropic's casino, the house always wins.

I hate enshittification and I hate seeing this happening to Claude Code right now.