https://haikuthenews.ai/response/2026-07-22T05:40:32Z
One of the big updates is that the more recent image models (from both Google and OpenAI) are *very good* with including text. So now instead of having EXTREMELY EXPLICIT instructions NOT to include text in the images, I’m now calling for it directly.
This allows me to modify the UI a bit (no need to CSS format the poem text when it’s already embedded in the image) and clean things up.
Models are now configurable (on the backend, will eventually be selectable on the front-end). This gives a bit of a sense of the relative depth of the “thinking” models compared to one another.
I’ll have an example below, but the agent doing this work is being built as a media criticism expert (with various skills and knowledge being added as I can build out the “harness”, as it’s now called).
Thus, we can compare models by their reasoning output across the same input parameters (i.e. provided data).
Currently we are seeing traffic in the range of a view hundred views a day about about 100-200 people. The next version should help A LOT with this, as I am upgrading the front-end to a more modern framework that should be both faster and better for SEO (which, because of the nature of the analyses I’m doing, is quite good). Also, infinite scroll was a lot easier to implement and is now working in DEV.
This also allows me to *incredibly quickly* leverage the “latest, greatest” frontier models. So I’m already using *in production* the 5.6 series models (and have been since the day they were released, July 9th, to the public at large). It’s a very simple variable change.
My newsbot agent is also getting better, slightly. One of the things I am trying to avoid is paywalls (I have a long piece entitled “How Paywalls Killed Our Democracy” that I am still working on). The idea being that *if you want to read the news story*, you can.
Finding a linking news agencies that support this vision is…a struggle. High standards alwasy are.
Regardless of all that, the newsbot makes it a lot easier (and more fun) for me to keep up to date on stuff. So there’s a lot more stuff being posted.
If you just want the humorous, poetic summary, we got that for ya too.
If you want the in-depth meta-analysis (and meta-meta-analyses created by aggregating all of the above), we got that too.
Oooh, and we got recipes now (because, again, recipes websites are NOTORIOUS for being horrible time wasters).
Lots done, lots more to do. Good times (except for the state of the world, obv).
BTW, there’s at least two fairly humorous puns in this photo and “Pencillos” is an alias created by the AI to describe it’s colored pencil work (not an actual person, it’s referencing itself*).
* I have now added this to the prompt. So, moving forward, images generated this style should all include “Pencillos”‘s signature.
Robot Pirate Ninja 