Claude is the best AI. And the worst
There’s a moment that comes to everyone who works seriously with artificial intelligence—a moment when you realize that the most impressive technology isn’t necessarily the most useful. For many of us, that moment arrived while using Claude.
On paper, Claude should be the obvious choice for anyone attempting substantial intellectual work with AI. Among the mainstream conversational platforms—ChatGPT, Gemini, Grok, Llama, and the rest—Claude consistently demonstrates the most sophisticated reasoning, the most nuanced understanding, and by far the best writing ability. It’s the system that feels most like collaborating with a thoughtful colleague rather than issuing commands to a sophisticated search engine.
Yet after months of attempting serious analytical projects with Claude, a troubling pattern emerges. The very platform that excels at complex reasoning becomes actively hostile to the kind of sustained work that such reasoning should enable. Claude works beautifully for cheating on your homework, resume polishing, and casual inquiry. For anything deeper—the kind of extended analysis that takes advantage of its genuine capabilities—it becomes not just inadequate but counterproductive.
This isn’t a story about technological limitation. It’s a story about corporate choices that transform excellence into frustration, and what happens when a company builds sophisticated tools but shows no interest in helping people use them seriously.
The Promise: What Claude Does Right
Anyone who has worked across AI platforms quickly recognizes Claude’s distinctive strengths. The differences aren’t subtle—they’re immediately apparent to anyone attempting anything beyond routine queries.
Claude writes like a human being. Not just competently, but with genuine fluidity and voice. More importantly, it can absorb your style preferences and strategic concerns without them feeling artificially tacked on. Other platforms produce serviceable prose; Claude produces writing that feels authored rather than generated. The distinction matters enormously for any project requiring sustained written output.
It actually understands what you’re trying to accomplish. Beyond literal instruction-following, Claude demonstrates remarkable ability to infer user intent and work with incomplete specifications. It can take vague requirements and develop them into coherent plans. It can work productively with ambiguous goals and evolving needs. This inferential intelligence makes genuine collaboration productive and pleasurable in ways that other platforms simply cannot.
It maintains intellectual honesty. Claude acknowledges uncertainty without becoming useless. Instead of fabricating confident responses or refusing to engage, it identifies exactly what it doesn’t know and explains why that limitation matters. This creates reliable boundaries for serious work—you can trust Claude to signal when you’re moving beyond solid ground.
It can think across domains without losing coherence. Complex analysis often requires drawing connections between different fields while maintaining rigor in each area. Claude excels at this kind of integrative thinking in ways that other platforms struggle with. ChatGPT frequently loses logical consistency across extended reasoning chains. Gemini tends to produce shallow connections between concepts. Claude can maintain both depth and breadth through complex analytical sequences.
These capabilities should make Claude ideal for serious intellectual work. They represent exactly the kind of sophisticated reasoning that could enable genuine AI partnership in research, analysis, and creative projects. The technology is there. The potential is obvious.
And that’s what makes the reality so frustrating.
The Problem: When Excellence Meets Indifference
The fundamental barrier to using Claude for serious work isn’t technological—it’s corporate. Every limitation that makes sustained analysis impossible with Claude stems from deliberate choices by Anthropic that prioritize casual users over serious ones.
You can never know where you stand. Unlike other platforms that provide at least basic usage information, Claude operates behind complete opacity. You cannot see your token budget, cannot track consumption, cannot estimate how much work you can complete before hitting limits. Anthropic tracks this information internally but deliberately hides it from users. This makes planning any substantial project impossible—you’re flying completely blind.
The rules change without warning during your work. Anthropic manages server load by dynamically adjusting limits and context windows, often while you’re in the middle of important conversations. A project that was proceeding normally can suddenly hit restrictions that didn’t exist an hour earlier. Worse, if Anthropic reduces context window size while you have a complex conversation in progress, you may be locked out of continuing that work for days. No other major platform implements such disruptive resource management.
When you hit limits, it’s a brick wall. Claude doesn’t degrade gracefully—it shuts down abruptly with no possibility of recovery. You cannot save your progress, cannot ask for summaries of what you’ve developed, cannot transition smoothly to a new session. Hours of analytical work simply becomes useless. Other platforms provide warnings or preservation options. Claude offers nothing.
The system cannot help you navigate its own limitations. Perhaps most frustratingly, Claude has no awareness of its operational state. It cannot tell you how much capacity remains, cannot advise whether to continue or start fresh, cannot explain why its performance might be declining. This blindness makes strategic collaboration impossible—you cannot optimize your approach because neither you nor Claude knows what constraints you’re operating under.
The company doesn’t want to hear about it. Anthropic has systematically eliminated feedback channels. No phone support, no email contact, only a GitHub repository that receives no meaningful review from company employees. Problems documented months ago receive no response. The message is unmistakable: Anthropic has no interest in users attempting serious work.
The interruptions are arbitrary and broken. Claude rarely warns when approaching limits, and the enforcement system is so poorly implemented that users commonly wait hours to resume work, only to be told they must wait again before processing a single new prompt. The desktop applications make things worse by automatically deleting any work in progress when you return from interruptions.
Claude cannot reliably maintain corrections across conversation turns, regardless of their nature. Despite billions in development investment, Claude fails at basic instruction persistence that should be fundamental to any collaborative system. Whether correcting factual errors, clarifying context, adjusting tone, correcting text, or specifying requirements, users cannot trust that Claude will remember these corrections in subsequent responses. Instructions that should become permanent modifications to Claude’s understanding instead get randomly ignored or forgotten, forcing users to repeatedly re-explain the same points. This creates a maddening cycle where sophisticated analytical work gets derailed by Claude’s inability to maintain basic conversational continuity. The system can perform complex reasoning but cannot execute the fundamental collaborative task of “remember what I just told you and apply it going forward.” This represents a stunning disconnect between technical sophistication and basic operational competence—like building a research partner with brilliant insights who suffers from selective amnesia about everything you’ve taught them.
Every one of these problems could be solved with straightforward engineering work. Usage indicators, graceful degradation, session preservation, operational awareness—these are basic features that other platforms provide. The technology exists. The solutions are well understood.
Anthropic simply chooses not to implement them.
The Larger Picture: When a Sophisticated Tool Meets Corporate Indifference
An substantive analysis reveals something specific about Anthropic’s approach to its users—a level of indifference to serious use cases that stands out even in an industry not exactly known for customer focus. Anthropic has created the most sophisticated conversational AI available, then wrapped it in service architecture designed to frustrate anyone attempting substantial work.
This pattern reflects a deeper misunderstanding of what serious AI collaboration requires. Building impressive technology is only half the challenge—the other half involves creating infrastructure that enables sustained engagement with that technology. Anthropic has excelled at the first part while showing active indifference to the second.
The result is a profound waste of technological capability. Claude offers reasoning sophistication that should enable breakthrough analytical partnerships. Instead, it functions primarily as an elegant tool for casual queries and brief assistance—exactly the applications that require no sustained engagement or strategic planning.
Users who recognize Claude’s potential find themselves in an impossible position. The platform that best supports complex thinking cannot support complex projects. The system most capable of sophisticated analysis cannot maintain the continuity that sophisticated analysis requires.
This transforms Claude from a tool for serious work into a demonstration of unrealized potential—impressive in individual moments, but ultimately unreliable for anything that matters.
The irony is sharp. In an era when artificial intelligence increasingly promises to augment human capability, the most advanced conversational AI becomes useful primarily for tasks that require no augmentation at all. Claude excels at the kind of thinking that should matter most, then systematically prevents that thinking from developing into anything substantial.
For those of us who have glimpsed what genuine AI collaboration might look like, Claude represents both promise and a uniquely frustrating missed opportunity. Anthropic has built the technology for transformative intellectual partnership, then systematically prevented that partnership from developing. The most advanced conversational AI becomes useful primarily for the most casual applications, not through technical limitation, but through deliberate corporate choice.
Claude stands alone among AI platforms—not for its technical excellence, which is genuine, but for the systematic way that excellence has been undermined by a company that appears fundamentally uninterested in supporting the serious work its technology makes possible.
