David Hobbs

This is a great example of where the LLM (just like a junior programmer might) was starting down the path of implementing a change for a particular *page* and I had to redirect it to implement a more generic, long-term approach.

When thinking about how/if/when to use LLMs for context (especially at scale), consider these cost drivers. hobbs.link/46rGrYK

I like the title from The Economist: Can you make it to the end of this column?

It's counterintuitive that the friendliest way to name an invite between you and one other person is Your Name / Their Name (like "David Hobbs / John Doe: Redirects") so that on their calendar they don't see their name first. But if *feels* friends to defer to their name first.

Let me know if you need a redirect audit. I've been doing some really interesting crawling / graph db analysis for a client as part of a replatforming / relaunch.

Those times when you think you've written something golden that's going to get lots of traction online and... crickets.

Agencies, when you are responding to an RFP for a redesign: if you had magical x-ray vision about the potential site(s), what would you like to know about the site? In particular, what would you most like to echo back to the potential client to show you understand their site?

Remember these are two things:

* Chat as a UI
* The underlying primary source of what is displayed

Yes, the ChatGPT app combines both.

But these do not need to be tightly bound.

Other options: Chat with data, or Graphical UI where the primary source content is an LLM.

The "see link in first comment" is kind of ridiculous. But I think I'm going to start doing that, especially on LinkedIn, to get more reach.

I sometimes add a Do Not Do list to implementation strategies. This can be useful when a team is losing focus, for instance by chasing a current trend that is not relevant to their current situation.

Yeah I think I'll mostly be doing ultralight CW SOTA / POTA. We'll see. But first I have to get my license after letting mine lapse a long time ago.

90% of situations: Need big web changes → Send RFP
10%: Need big web changes → Define/align internally → Send RFP

When needed:

* Changes spanning silos
* Restructuring cross-system integrations
* Competing internally desires

Sending an RFP already locks many things in.

Thinking about getting back into amateur radio. Transmitting in Hell Mode is not something I remember.

Every web team needs some systems thinkers.

Announcing the Official PHP SDK for MCP The PHP Foundation — Supporting, Advancing, and Developing the PHP Language

From the PHP Foundation: Announcing the Official PHP SDK for MCP hobbs.link/47CdHPj

For Pete's sake, stop the popups that ask for my email address to get a discount only to surprise me with a request for my phone number to actually get the discount.

Use Cases for AI in Digital Presence and Content Analysis | Articles | Content Analysis Database LLMs open up new and exciting opportunities in content analysis

I've been searching for solutions to some use cases that newer LLMs finally help resolve. Here's some Use Cases for AI in Digital Presence and Content Analysis: hobbs.link/41HywVI
Note: LLMs are not a universal cure and should only be used where it makes sense.

Would love to normalize prominently displaying what you're about to be charged, and not the "monthly" fee if you pay for a year. Show me the monthly price if I'm paying monthly, and the annual price if it's an annual subscription. Lead with that.

I miss breadcrumbs

I question time-on-page as a useful engagement metric. If someone can come to your site, find the info they need (or accomplish what they need), and leave quickly then isn't that ideal? Have there been any studies looking at whether time-on-page actually captures engagement?

The *class* of tool is probably more important than the particular tool.

A useful LLM use: walking through a website and just talking out loud about the problems/opportunities then having the LLM summarize. The bit I haven't worked out is how to automatically capture URLs of the pages I'm looking at.

When it comes to defining xpath/regex, I've found a useful feedback loop:

1. Give a stab at writing the pattern
2. Use the pattern to extract from some content
3. Ask an LLM why the pattern didn't work against that content

Repeat.

You can ask an LLM to extract/summarize from content. At scale, this can work but it's slow and expensive. Or you can ask an LLM to define a pattern (for example xpath/regex) and then quickly and cheaply scrape it out.

In the end, you still need to define what you want to accomplish. And someone needs to be able to evaluate whether the approach (or code, or organizational processes, or tool, etc) has a chance of pulling that off.

One of the biggest things to define before sending out an RFP for web redesign/improvement: what is the extent of changes (which websites) and which backend systems will be impacted?

Content Chimera Analyze, understand, and frame complex sites from chat

Sign up for the Beta: hobbs.link/2GKRMs8

Once invited: Enter URL → Chat with content inventory

Content Inventories with AI

Thanks Tim!

Hmm, I use em-dashes a lot— now I hear this can be a marker for ChatGPT output?

That moment when you're not sure how a bit of code ever worked.