[ClaudeAI] Prioritize core model quality and reliability over new features

Hi. Claude Max user here. First, I want to acknowledge the work that’s gone into Claude Code. I appreciate the effort. But this is a serious criticism aimed at leadership and the product team, because I’ve spent hundreds of dollars on Claude subscriptions and I’m not getting the level of service I’m paying for. 1. Model quality and reliability have clearly declined. I’m not going to restate every complaint people have already been making, but I want to add my voice to that pile. I’ve experienced the drop in model quality myself over the past few weeks. The reliability hasn’t been great either. Your status page shows 98.73% uptime. A single nine. For a tool developers rely on for production work, that’s unacceptable. 2. You have a serious compute constraint problem. From the outside, it seems obvious that demand is growing faster than your GPU resources. And the product decisions reflect that. Tighter usage limits, less generosity after outages, nudging users to clear context to reduce token load, restrictions on third-party access, and other signs that you’re trying to stretch limited capacity. I understand why you’d do that. But let’s call it what it looks like. 3. I actually do understand the pressure you’re under. This is not coming from someone who thinks scaling an LLM product is easy. I understand that explosive growth, infrastructure bottlenecks, and the realities of running a non-deterministic system at scale create real engineering and operational problems. I get that. And even though I’m paying a lot, I’ve tried to be patient. I’ve tried to overlook the rough edges because I assumed the team was working through them and would stabilize things. 4. What I do NOT understand is the product strategy. This is my main frustration. Why are you shipping "fluff" features while the core engine is smoking? That’s the part I genuinely do not understand. Why are leadership and product pushing out something new every other day when older features still have obvious bugs, reliability issues, and support gaps? 5. The "/buddy" thing is exactly the wrong signal. A few days ago I opened Claude Code and saw the colorful `/buddy` command. Apparently it’s a little creature that sits in the terminal and comments on your interactions. Sure, it’s cute. But I don’t know a single developer who was asking for a terminal pet while the core experience is getting less reliable. More importantly: **This is an unnecessary prompt hitting your already-strained GPUs**. Why waste compute on a gimmick when your primary models are struggling with latency and reliability? That’s what makes this so frustrating. When users are already feeling slower responses, tighter limits, degraded quality, and instability, shipping novelty features sends the message that priorities are badly misaligned. 6. Most users would trade "more features" for "works every time." I like seeing products improve. I like seeing active development. I like useful new features. But I’d rather have a car with just a steering wheel and four tires that starts every morning than a luxury sedan with a 15-inch Dolby touchscreen that breaks down on the highway. We are here for the intelligence and the uptime. If the car doesn't drive, the buddy in the passenger seat doesn't matter. Right now, it feels like too much energy going into extras, not enough going into making the core product dependable. And to be blunt, a lot of your current users (maybe even myself) are paying because Claude is still one of the best coding models available. But the moment there’s a better coding model with better reliability and a better overall experience, many of those users will switch immediately. Please fix the core, then ship the toys. I like Claude, and I’d like to keep liking it.