I did the same marketing work every week for 10 years. So I built AI systems that do it now :D
Netherlands
Joined August 2010
- Tweets3.2K
- Following317
- Followers797
- Likes2.2K
Before you build your next tool, ask yourself one question: what problem does this solve?
There is a useless tool epidemic going on in most companies now.
I love nothing more than seeing my PPC friends making awesome tools. Many of you also ask me to review them or give my opinion, but sadly in the past couple of months I had to say more often than I'd like that the tool doesn't solve the actual problem.
Many of you now make fancy dashboards, ad making tools or other in-house solutions to "help" your work or your clients' work. But most of the time the actual problem is still there, we just procrastinated and built something fancy instead.
Let me give you two examples:
1. A dashboard to "understand the data better"
I talked with a marketer who was facing growth problems and couldn't reach the targets. So he spent 2.5 days making a super fancy dashboard to get a "better" overview of the data, hoping the solution would show up.
Not to mention they already had Power BI and a Looker dashboard showing more or less the same, just in a less fancy way.
They came to the conclusion, it looks like the problem is the conversion rate. Which you could already see in Looker :D was literally the 2nd card :D
So instead of working on the landing page and the actual problem, now we have a new dashboard.
2. A feed tool instead of a feed fix
Another case where the feed had issues, attributes were not filled in correctly and a lot of data was missing, so filling them in was simply not possible. So now there is a tool where you can load the products, filter on missing attributes and then use AI to fill them.
The 3 days that went into fighting the API to get the products loaded correctly, I would have spent on scraping and gathering the missing data. Once you have that, filling the feed is a loop in n8n you build in 30 minutes.
Now that building is easier than ever, your time is even more valuable. Spend 30 minutes more thinking about what the actual problem is before you jump to build something, because building feels like solving, but most of the time it isn't.
AI is a great helper, but for people who like to procrastinate it's also a huge danger.
I think we already have AGI/SI and we just don't realise it. Claude Opus 5.5 made an 8 min video based on two simple prompts.
So last week I visited NYC for the first time in my life. I usually take a lot of videos, and when I get home from a trip like this I make an edit that I keep. Normally I spend a weekend cutting a video like this myself, because it's the thing I keep from a trip.
So I thought, why not ask Claude to try it?
Claude only had one thing: a skill I made which instructs Claude to use Remotion for video editing. I added the raw conversation so you can see how little input I had to give.
And the output is really awesome. I mean, it's not perfect, but holy shit, how good these models are.
Here is a short summary of what Claude did:
1. It found out only 3 of the 949 files were on my Mac and pulled the other 10 GB down from iCloud
2. It used the GPS in the photos to work out what we did each day
3. It made contact sheets of every shot, went through all of them and picked 180
4. It wrote a story: landing at JFK, the awards night, the ferry, the Nets, Empire State at sunset, flying home
5. It generated original music with ElevenLabs
6. It rendered it, checked it, and made a smaller copy for family that I never asked for
It even turned the music down at the part with the 9/11 Memorial footage, insane.
I added the video in the next post, so check it out. It's not perfect, but these models can do a lot more than we think. I think this is also almost AGI-like.
With such low input, he really figured out a lot on his own.
@bcherny you guys are insane.
There is a lot of AI music flooded into YT Music, and I mean a shit ton.
I loved the Discover feature of YT music, but now it's unusable.
I have nothing against AI music if it's good, but most of them are shit, just shit.
After spending a few days in NYC is really like spending time in a movie, but I would not be able to live there.
I think this mystique wears down quiet fast :D.
But 1 thing NYC and US in general has that I envy is breakfast places. I love the diners and their wide range of breakfast options.
In NL weekdays be happy if you find something open at 7am to buy, let alone eat an awesome breakfast.
Also, I didn't know there are so many Spanish speaking in NYC. I think I heard more Spanish than English.
Been to NYC the only thing I was not able to understand is why do people stand in line for a bagel/pizza? Explain please?
I went one block further and had one of the best bagels I had and I didn’t have to wait at all.
What’s the lore?
I have to admit after to days in NYC this city is awesome.
I think I never took so many photos of staircases in my life 😅😂
But when I see one all I have is movie flashbacks
Are governments paying influencers nowadays?
There are a few people on this app that mostly post about why you should relocate to certain country and how good life is in certain country.
One post is not a big deal, but these accounts post about it like 4 times a week.
Most ships crashed and we never heard of them.
A few of them made it now we focus on those
Indie hacking exactly like this 😂 we only hear of the ones that made it which is a small percent
What is this cringefest going on between the OpenAI and Spacex people doing this tweet fighting instead of shipping?
I had to mute a lot of people already
Any PPC friends in the NYC area in the upcoming days or at the US Search Awards?
I’m happy to have a chat or coffee! 😁
Really well explained why I pay €2.45 per liter now
youtu.be/OETnuwwsv9U?is=K2DO…
While working on updates and the new version of PPC OS I can't help and constantly think of this picture.
What is that I can make simpler and better compared to how it is now.
It's the only way forward
It's weird that the same communication techniques work with agents as humans.
Agents states/ask something.
Me answer and also add a line to confirm I understood him correctly.
Saved my ass way too many times :D
Can't wait to test this, but the stupidly interesting thing is the 1M token OUTPUT limit, that is fucking insane.
For example Fable and Astra cap out at 128K
This does mean bigger room for reasoning, super interested how the quality will be on this.
With a 1M token output limit, Argon adds a deeper level of reasoning to tackle long, multi-step problems in one go.
Feedback from early testers will help us strengthen our systems before we roll out more broadly to developers, enterprises, and consumers soon.
Find out more → goo.gle/4rZBiSd
So while the 9 Games trend going, I made a small site that gives you an explanation on what the games tell about you.
According to my card I should work at IKEA :D
Get yours at : player-card-alpha.vercel.app
