AI Beyond the Hype

As the AI hype cycle cools, the real question becomes: what is this technology actually useful for?

I believe that even if venture capital dries up and model progress plateaus, AI will remain extremely useful — just not in the ways most people expect.

Let’s get to the point.

I suggest it won't be anything related to what we think about when we hear "AI".

I’m not claiming to have all the answers, but I can offer a glimpse of where I think this is going.

I’ll illustrate this using my own agentic system, Contenox, which I’m developing from scratch in Go.

__

Here’s a small but concrete example.

I still need a couple of attempts to get the engine running, encountering errors like default-provider is not set. or responses such as "I don’t have a plan-manager tool available in this environment, so I can’t literally invoke it. But I can provide the plan it should contain."

For context, Contenox began as a workflow engine for infrastructure and governance tasks, so it still leaks some raw engine details, but after a couple of attempts poking the system, I was able to get the vibe right.

__

It’s challenging to keep up with everything happening in tech when most of your time is spent on a day job... or just daily life. So I try to automate as many recurring tasks as possible, such as tracking dependencies used in my projects.

It's quite simple and I'm certain any MCP compatible agentic system can pull this off. You just need to initially register the tools Playwright and Notion before prompting.

contenox plan new "use playwright to go to https://github.com/ollama/ollama/releases and document the changes in the past 5 releases into notion using the notion tool" --shell

The prompt explicitly tells the system which tool to use for each task. This prevents the model from, for example, trying to open Notion with Playwright or using web search instead of the browser.

In Contenox, a plan consists of multiple steps the model believes are necessary to achieve the goal. This is a common concept in many agentic systems

Generating plan for: use playwright to go to https://github.com/ollama/ollama/releases and document the changes in the past 5 releases into notion using the notion tool...
Created plan "plan-9d3ef440" with 6 steps. Now active.

Trusting GPT-5.4 here to get it right with the auto mode...
(It's an awesome model for tool related tasks btw)

contenox plan next --auto

Executing Step 1: Open https://github.com/ollama/ollama/releases in Playwright and wait for the releases list to fully load....
✓ Step 1 completed.

Executing Step 2: Identify the 5 most recent releases shown on the page and, for each release, capture the version tag, release title, publication date, and release notes text....
✓ Step 2 completed.

Executing Step 3: For each of the 5 captured releases, extract and summarize the notable changes into concise bullet points, preserving any clearly labeled breaking changes, new features, fixes, and platform-specific updates....
✓ Step 3 completed.

Executing Step 4: Create a new page in Notion using the notion tool with a title such as 'Ollama GitHub Releases - Last 5 Versions' and include the source URL https://github.com/ollama/ollama/releases near the top....
✓ Step 4 completed.

Executing Step 5: Add a section for each of the 5 releases in the Notion page, including the version tag, release title, publication date, link to the specific GitHub release, and the summarized change bullets....
✓ Step 5 completed.

Executing Step 6: Review the completed Notion page to confirm all 5 releases are included in reverse chronological order and that the summaries accurately reflect the GitHub release notes....
✓ Step 6 completed.

Executing Step 10: Review the Notion page content for completeness and formatting, then save and confirm that all 5 release summaries are present....
✓ Step 10 completed.
All steps complete. Plan is done!

Done!

The example is simple, but the pattern scales dramatically. Apply the same workflow to generating a full user manual...
...capturing screenshots with Playwright, organizing sections automatically, and publishing the result into a structured Notion document.

To wrap this up I'll show the chain of steps Contenox did:

contenox plan show
Plan: plan-9d3ef440 (active) — 6/6 complete
1. [x] Open https://github.com/ollama/ollama/releases in Playwright and wait for the releases list to fully load.
2. [x] Identify the 5 most recent releases shown on the page and, for each release, capture the version tag, release title, publication date, and release notes text.
3. [x] For each of the 5 captured releases, extract and summarize the notable changes into concise bullet points, preserving any clearly labeled breaking changes, new features, fixes, and platform-specific updates.
4. [x] Create a new page in Notion using the notion tool with a title such as 'Ollama GitHub Releases - Last 5 Versions' and include the source URL https://github.com/ollama/ollama/releases near the top.
5. [x] Add a section for each of the 5 releases in the Notion page, including the version tag, release title, publication date, link to the specific GitHub release, and the summarized change bullets.
6. [x] Review the completed Notion page to confirm all 5 releases are included in reverse chronological order and that the summaries accurately reflect the GitHub release notes.

Yep that's it.

It's an interesting period to transition from coding a project to actually using it. A bit dull, a bit boring, but a very necessary step.

I’ll keep you posted. Hopefully it won’t take long to hide Contenox’s raw engine behind a user-friendly UX.

This is what “AI beyond the hype” may actually look like: not artificial intelligence replacing humans, but reliable systems quietly handling digital work at scale.

Cheers!