Post #4747681 · Elektrine

2026-09-18 11:12 UTC

From this point on, when I write “AI,” I primarily mean systems based on large language models. LLMs are tools we have barely begun to understand. Rapid progress in agent harnesses, including recent work around DeepSeek Harness, makes that hard to miss. Context, memory, tools, orchestration, and verification can radically change what the same underlying model can do. As Kristian Köhntopp argues, using an LLM for complex work without local knowledge, clear instructions, intermediate steps, and review is like judging a new employee before onboarding them. Then blaming the employee for not knowing where anything is kept. Before embracing or rejecting LLMs, we should first understand what sort of tool is sitting on the desk. False promises do not help. When they fail, they harm people and make responsible adoption harder. 4/34

Replies (1)

@masek@infosec.exchange 2026-09-18 11:13

One distinction is essential: training is not inference. During training, a model learns its parameters from enormous datasets through repeated computation across many accelerators. This is where most upfront demand for compute, energy, and training data arises. It is also where many questions about how content was acquired belong. During inference, an already-trained model applies those parameters to a new input. Each run still uses energy, and usage at global scale adds up, but it does not repeat the training process. It is far less resource-intensive per task. Efficient runtimes and quantization now make surprisingly capable inference possible on home hardware. Qwen3.8-27B is one example. Training and inference belong to the same system, but not to the same bill. Mixing them together makes the discussion simpler, which is not always the same as making it useful. 5/34

Open ##4747680