LLMs Are Still Toxic, Stuck in the Past, and Bad at Math

The author ran 200 addition problems on GPT Sol High and it missed one. The model doesn't calculate—it predicts the next likely digit. ChatGPT gets it right because a harness hands the problem to a Python script. The post walks through the same pattern for three other unsolved flaws: stale knowledge patched by RAG, limited context windows, and toxicity still baked into the model. The real progress isn't in the models but in the tooling wrapped around them.

Why it matters: A developer-perspective long-read with experiments and sharp judgments, dissecting why LLMs' four old flaws (math, staleness, short memory, toxicity) persist and arguing progress came from tooling, not the model. Hits all three HKR axes, but as a commentary/survey rather than ...