Act now. Adjust later.
piotr(at)grnd(dot)pl
Copilot → Cursor → Claude Code → ???
The question is not "which tool?"
Are you building the muscle to use any of them well?
The technology was ready
before you were.
The bottleneck is not silicon. It's everything else.
The single most underestimated element in enterprise AI adoption.
01Same question. Six strategies. Differentiated entirely by context.
A frontier LLM with a bare-bones prompt is a world-class chef with no pantry, no guest list, and no idea if the kitchen has an oven.
AI slop doesn't come from bad models.
It comes from starved models.
There is no such thing as a second first impression.
One context-starved demo → stakeholder lost for years.
The corporate equivalent of eating your vegetables. Nobody's favourite topic. Transformative over years.
02Little reward for doing it well.
Almost no visible punishment for neglecting it.
Until you try to make your enterprise agentic and discover the agents have nothing to work with.
Your greatest asset and your biggest obstacle — simultaneously.
03Every efficiency, every shortcut, every "I just know how to do this" — neural pathways carved over years.
Renovations always take longer and cost more than the estimate.
"What will you do with the two and a half weeks you just freed up?"
Warm it well before serving.
04Think kitesurfers, not container ships.
Terrible endpoint. Great waypoint.
200 agents = 200 executable intents. Encoded context. Defined workflows. Real problems with working answers.
You don't plan base camp to peak on day one.
Generative AI is non-deterministic. The same input can produce different outputs.
Build feedback loops. From day one. For everything.
The winners are not those who get it right first — but those who get it less wrong, faster.
None of this happens
by accident.
Intentionality is the difference between an organization that becomes agentic and one that merely buys agentic tools.
Tactical plays grounded in principles. Pick what fits.
05A "skill" tells an LLM how to do something well. Skills are the new currency of productivity.
Personal agentic tools — OpenClaw, Hermes Agent — are how people master context engineering.
Models on consumer hardware now exceed 2024's SOTA benchmarks. 256K context windows on a $1,500 gaming PC.
A well-crafted CLI facade can replace a constellation of middleware, MCP servers, and integration platforms.
~$100/month per person. Let them pick their own pickaxe.
If I had to pick one agent to build first, it would be this.
What comes next is a Knowledge Manager — one whose job is not to answer questions about your processes, but to run them. That's the Executable Intent.
Better to have an imperfect Corporate Brain today than a perfect one in 2028.
Your entire org already has access. That's not a limitation — it's a launchpad.
It's no longer text only. Video summaries. Podcast briefings. Interactive dashboards.
As people master these tools, they become extraordinarily attractive to the market.
Not because the technology demands it.
But your competitors won't wait.
Neither will your best people.
One cup at a time, warmed well.