11 years ago today, I took my first uber. I was so excited that they'd arrived in Perth that I posted to Facebook about it.

Man, when did Adobe start installing so much malware with its applications? My PC barely runs now because of all the background tasks, services and scheduled tasks.

Signed up for Gemini Pro this week. It hijacked the AI assistant on my phone and my directions took me to "Karrinyup, Washington". (WA = Western Australia). Worse, my phone's travel directions are now polluting my Gemini history.

Spent a bunch of time the last couple of weeks making changes to my Recipe Engine that allow me to more quickly go from "Catalogue Published" -> "Recipes published", basically brings it down to a few hours. Now I can focus on growth, and better recipes and images #buildinpublic.

And this week's recipes for @catalogkitchen are up! . ~21 meals to feed 4 for ~$20 (or less!) based on this week's @Coles specials. #CookingAustralia #AustralianFood #Perth #PerthFood

So, rather than fixing the whole pricing calculation process, I think I can just start green-flagging things that are guaranteed to be right, and then it'll make it a bit easier to distinguish between "probably right" and "weird edge cases".

The best part is it grows over time. If I find myself always matching to the same ingredients, week after week, I just add it to the list of auto-matches. A task that takes an hour now takes 5 minutes.

Certain ingredients pretty much always match to a particular ingredient in a store, so I just auto-match those and double check the pricing is best. It gives me a quick visual queue that lets me quickly scan past matched items and reduced my manual workload by about 80%

I had a similar problem when I was doing the ingredient matching - matching "onion" to "Coles Brown Onion 1kg prepacked" as the best culinary match at the best per unit price. I managed to optimise this in a few ways, most recently by adding a green flag system for known matches.

So I've got a bit of a solution to it, but it means rebuilding this part of the engine from scratch. Based on learnings, of course, which is great. And to make the whole manual workflow easier. But every minute I spend doing that is a minute I'm not focusing on user acquisition.

And we're talking weird shit, like logic to handle "0.5 of 1xcabbage each" when "1x half cabbage each" is the sale price (they're both sold as "each"), or how many sprigs of parsley in a bunch. Even if I hand off these to a decision engine, there's still weird edge cases.

The problem is that while I've automated a lot of logic for recipe generation, and price/macro calculation, there are hidden edge cases that mean I have to review pricing for each ingredient, in each recipe, in reach region, EVERY week. Very time consuming.

Every week I hit the same wall with @catalogkitchen:.“Just push the recipes and focus on marketing”.vs.“Fix the stuff that makes pushing them unbearable”.I'm so torn between the two, because I know user acquisition is so critical at this point.🧵👇.

"There's no ozempic for personality" is one of the funniest things I've heard this year. Thanks @garyvee.

RT @robertherjavec: Entrepreneurship isn’t a hobby, it’s a crucible for character. Every missed event deepens your grit and sharpens your….

I'm going to launch a new trend called VibeDoctoring. You just punch in your symptoms and do whatever it tells you. Sure, we might kill a few people along the way, but just think how much time and money we'll save by not having to wait for real doctors.

Took like 20 edits in ChatGPt and then I just had to photoshop the one that looked closest (but not best) to what I wanted.

I made a thing.(honouring hoppy doodle's old pic)

So, easy to write code to fix this. But it calls every other conversion it's doing into question. Especially where I have a defined output format (json function call schema in OpenAI API), I shouldn't have to double check that the results are correct.

"olive oil1.5tbsp = 12.6ml". It's stuff like this that really underscores the unreliability of AI. I've asked it to convert from a Recipe Unit to a Base Unit (g for solids, ml for liquids). Converting a solid from tsp to g would be difficult, but tbsp to ml should be easy.