A blip appeared on my phone while I was on the treadmill last week, it was yet another LinkedIn post talking about the wonderful value AI has brought upon us. Day after day, anywhere I look, I have to sit through opinions that are exuberantly optimistic. Yet all examples given are always tiny tools, and I have yet to see a real-world example of agentic usage in production critical systems that hasn’t gone wrong, or at best underperforming.Don’t get me wrong, I derive extensive value from the use of AI and I do enjoy using coding agents for prototypes, tools and debugging. But I sense a disconnect between what the AI evangelists say, and what I experience first hand. We need some pragmatic counterbalance to ground some of the marketing fluff in reality.So here are some of my hot takes on the current state of AI, March 2026.# Hot Take 1: Anything In Public Can Be Learnt If You Want ToWith AI Big Tech currently heavily subsidising our freely accessible usage of sequence guessing generators, everyone is able to to skip right past the data organisation and accessibility phases of organising the world’s information and making it universally accessible and useful to ourselves. The chatbots are now capable enough to do JIT search MCP, so we can worry less about recency of information.During my bus rides, I’d use Gemini Deep Research to do a deep dive on stuff like “Latest Trends in Incident Management, Anchored to March 2026” to keep up with the latest industry practices. This would provide me with adequate entertainment on my 22km commute, which is otherwise an inexplicably 1h40m long, excruciating bus journey home. If anything in the generated article feels incredulous, I would pull out my dusty Google-fu and fact check the information.Although… I can no longer tell if LLMs are becoming more accurate, or I’m fact checking less, or the LLMs are hallucinating more convincingly now.Some of my teammates have asked how I know so much about everything. The secret is to mak...
First seen: 2026-03-29 05:48
Last seen: 2026-03-29 13:53