Changes in AI - Part 1

The Last Three Months Changed Everything

If you blinked between February and now, you missed an entire era of AI.

That's not hyperbole. The pace at which artificial intelligence has shifted — from a tool we use to a collaborator that acts — has compressed years of expected progress into a single quarter. For those of us building, working, and living alongside these systems, the ground has moved under our feet.

From Copilot to Colleague

Three months ago, AI was still mostly a really good autocomplete. You'd ask it something, it would answer. You'd give it a draft, it would polish it. The relationship was transactional, turn-by-turn, with you firmly in the driver's seat.

That model is already obsolete.

Today's AI agents take on multi-step tasks, navigate ambiguity, ask clarifying questions, and execute work autonomously across hours of effort. They write code that compiles, run tests, debug what fails, and ship the result. They read your codebase, understand its conventions, and propose changes that fit. The shift from copilot to colleague isn't a marketing slogan — it's the lived experience of anyone paying attention.

What Actually Changed

Three things converged:

Reasoning got real. Models now genuinely think through problems — planning, backtracking, verifying their own work — rather than pattern-matching their way to plausible-sounding answers. The gap between "looks right" and "is right" has narrowed dramatically.

Context windows expanded into actual memory. What used to be a goldfish-brained assistant now holds entire codebases, multi-document research projects, and ongoing conversations in working memory. The friction of constantly re-explaining yourself is fading.

Tool use became fluent. AI no longer just talks about doing things. It opens browsers, edits files, queries databases, sends messages, books appointments. The boundary between conversation and action has dissolved.

What This Means for the Rest of Us

For builders and operators, the implications are stark. Workflows designed around "ask AI, then do the work" are being replaced by "describe the outcome, then verify the work." That's a fundamentally different skill set — less prompt engineering, more delegation, judgment, and review.

For everyone else, the everyday experience of getting things done is quietly being rewritten. Research that took an afternoon takes ten minutes. Code that needed a specialist gets drafted by anyone. Strategy documents, financial models, marketing copy, technical analyses — the friction of producing high-quality work is collapsing toward zero.

This isn't always comfortable. It raises hard questions about expertise, employment, judgment, and what humans are uniquely for. Those questions deserve serious attention, not dismissal in either direction.

The Open iPub Premise

This is why Open iPub exists now, in this moment.

We're not here to hype AI or to fear it. We're here to think clearly about it — to track what's actually shifting, examine what it means for the work we do and the world we're building, and share the practical lessons of using these tools in the real world. From sustainability to climate action to the systems we rely on every day, AI is becoming an unavoidable layer of how things get done.

The next three months will bring changes we can't yet predict. The three months after that will bring more.

The goal isn't to keep up. The goal is to think well about what's happening, build with intention, and stay grounded in what matters.

Welcome to Open iPub. Let's figure this out together.