𖣂📈 The next evolution: IMR at inflection; DAI & the erosion of tacit knowledge; Goodbye Claude & hello Kimi!

HOMODEUS RESEARCH WEEKLY BRIEF · 18 August 2026 Research | Videos | Podcasts | LinkedIn | Follow on X When I grow up I want to be a prompt engineer… Hi, Back in February, I was eagerly learning how to “prompt properly”. Hacks with structured ways and key words such as “ACCURACY IS CRITICAL” to write prompts and speak in “AI-Native” such that the we could get LLMs to do what we want. Last week I cancelled my Claude account, now using OpenAI as my main and dabbling in Kimi. Looking back at some of the prompts I was learning to write earlier year compared to what’s required now provides perspective on how quickly AI has evolved. My prompt in Feb-26… Same prompt today… Is the AI getting smarter? Yes, but perhaps not in the way you think. The “smartness” is coming not from greater compute, but thru the transfer of our tacit knowledge to the LLMs. What is tacit knowledge? Quintessentially, it’s the IP of “our process to get a job done”. Some of this should be ceded to AI as it increases convenience, but there’s a fine line between convenience and when we start ceding what is industry standard vs. proprietary process (and thus knowledge). How can we optimise our workflow to optimise for…