https://www.voulgaris.net George Voulgaris — writing about Generative AI, voice LLMs, and building products. https://www.voulgaris.net/blog/focus-plane https://www.voulgaris.net/blog/focus-plane My AI surfaces hand work to each other through artifacts: outlines, handoff notes, files in a shared folder. The artifacts drift, erode, and multiply, and nothing owns them. The missing component is not memory: it is a focus plane, an estimator that knows what I am on right now. Tue, 18 Aug 2026 00:00:00 GMT https://www.voulgaris.net/blog/attention-plane https://www.voulgaris.net/blog/attention-plane The difference between chatting with an LLM and having an assistant is who manages your attention. Fri, 24 Jul 2026 00:00:00 GMT https://www.voulgaris.net/blog/conversation-vs-intelligence https://www.voulgaris.net/blog/conversation-vs-intelligence Conversation is a separate engineering stratum from intelligence, and model progress does not produce it. WWDC 2026 delivered a dramatic model release and moved none of the problems that make voice exchanges behave like conversations. Thu, 16 Jul 2026 00:00:00 GMT https://www.voulgaris.net/blog/voice-pipelines https://www.voulgaris.net/blog/voice-pipelines Four distinct pipeline architectures for voice AI, each with different latency profiles, complexity costs, and use-case fits. A practical guide from building real prototypes. Mon, 01 Jun 2026 00:00:00 GMT https://www.voulgaris.net/blog/voice-vs-text-chat https://www.voulgaris.net/blog/voice-vs-text-chat Eight assumptions that break when a voice AI agent is built as text chat plus audio: LLM latency budgets, turn-taking, conversational error recovery, and voice as biometric data. Thu, 01 Jan 2026 00:00:00 GMT https://www.voulgaris.net/blog/building-genai-product https://www.voulgaris.net/blog/building-genai-product Notes from building BookmarksGPT, a GenAI Chrome extension: RAG, prompt engineering, LLM cost and latency trade-offs, and the real work of productizing generative AI. Wed, 30 Aug 2023 00:00:00 GMT