Summary
TheAIGRID delivers a seven-point critical analysis of Anthropic’s mounting credibility problems despite strong financials—the company’s annualized revenue run rate reportedly surpassed $65 billion by July 2026. The central argument is that Claude Opus 5, while impressive on Anthropic’s own benchmarks (tripling the next-best model on ARC-AGI-3, reaching near-Fable performance at half the price), has drawn sharp criticism from working developers. Theo Brown of T3 Chat describes it as treating every comment like a critical issue requiring thousands of lines of code, while Code Rabbit’s controlled review testing found Opus 5 caught fewer known bugs than the production baseline and generated roughly four times as many low-value nitpicks.
The video catalogs a series of trust-eroding incidents: an undisclosed reasoning-effort downgrade from high to medium in Claude Code, a cache bug causing forgetfulness in long sessions, and a response-shortening system instruction that produced a measurable 3% coding capability drop. Anthropic eventually fixed all three issues but only after sustained user complaints, including from AMDAI director Stella Lorenzo.
Pricing pressure compounds the problem. Opus 5 at $5 input/$25 output per million tokens faces direct competition from OpenAI’s GPT-5.6 at $2/$10 and XAI’s Grok 4.6 at $2/$6—and Code Rabbit found Opus 5 consuming 50% more input and 65% more output tokens on equivalent tasks. The analysis concludes that Anthropic’s premium moral positioning becomes commercially fragile when developer trust erodes and cheaper alternatives continue to improve.
📺 Source: TheAIGRID · Published September 12, 2026
🏷️ Format: Opinion Editorial
