VP Martin

Futurist 🔮 | Developing AI Agents 👨‍💻 at Alayans

[VP Martin @VP_Martin1]

There is hope for life after "subsidized AI usage"!

Never been a big fan of OpenAI, but they are setting a very important trend here.

AI use cases over API cost have become viable again.

[VP Martin @VP_Martin1]

@stemonteduro Question to envoke a thought, how would DG differentiate from something like the thing Tibomaker launched today? The Reddit leads generator. Different usecase, but similar toolstack, am I right?

[VP Martin @VP_Martin1]

@rouvenlue @publicappscom @PatrickHaede Same here. I am currently debating the same. I asked my wife (non tech, outside of AI hype). Her reaction was, people even buy pre-cut vegetables...

People buy services, in this case, the service to not create/vibe a setup yourself.

[VP Martin @VP_Martin1]

@ui_jedi Nice! Same ELO! DM me your username, lets battle some time ;-)

[VP Martin @VP_Martin1]

@catalinmpit Just make peace with the fact that if you can think of an idea, 1000 of others can so as well, if you can vibe code an MVP or landing page for it, 1000 of others can.

But, actually bringing to market and make it the best product ever with customers. That is not

[VP Martin @VP_Martin1]

@todo_trev Most intetersting read in a long time! Who knows, maybe you spark an idea in Elon’s head, like that interviewer did once at SpaceX :-D

[VP Martin @VP_Martin1]

Creating video tutorials is a lot of work,
so I created an app that does it for me.

Problem:
Video tutorials are the best resource for users (nobody want to read a bunch of text) but creating tutorials of every feature of your app is a lot of work

and..

change the UI, make an update to the code and the video(s) have become stale..

Solution:
Prompt + URL = video tutorial

And it can monitor changes in the code to auto reproduce any video tutorial when needed.

A 'self creating/updating help center'

Does that sound interesting?

Let me know if you want to try it out yourself.

[VP Martin @VP_Martin1]

@notacritix NGMI

[VP Martin @VP_Martin1]

Perspective.

I am working from my mom's house today.

So I am here running 5 AI agents in parallel and still feel like 'I am behind' and not productive enough.

Meanwhile, I see my mom's boyfriend, a farmer, busy for hours scanning in invoices manually one by one.

We are moving at different speeds and honestly, I don't know which speed is the best one.

[VP Martin @VP_Martin1]

@stemonteduro I am even poorer, got only $85

[VP Martin @VP_Martin1]

@stemonteduro The level of brainrot is high on TikTok, damn! :-o

[VP Martin @VP_Martin1]

@Mulan_London_UK Hi! You make some great points. I have a similar view on how AI can improve a PM's workflow.

We're building something around this and looking for people willing to give honest feedback. Let me know if you're interested.

[VP Martin @VP_Martin1]

@Asta_M_Sonner It's like has been said a while now, taste is a great skill to have

[VP Martin @VP_Martin1]

Saw some AI slop in the wild today.

Funny how everyone can make 'nice' posters, images and websites now.

But 'nice' is now the baseline.

Years ago, the lowest effort poster was made with WordArt and MS Paint.
Today it's a one-prompt ChatGPT image...

The lowest effort website was a poorly designed HTML + CSS page.
Today it's a one shotted AI website.

Low effort, low skill will always be recognized, no matter how good the tools become. When everyone can create something great, great becomes just good.

[VP Martin @VP_Martin1]

Agentic assisted coding is becoming a costly joke.

(proof I wrote this post myself, AI would never come up with this line 😏)

Hourly cost of running a coding agent is matching or exceeded human hourly cost.. And the actual compute cost is still subsidized, yikes.

But there are always ways to optimize, one of those is using an orchestrator/minion setup for a coding task.

Orchestrator (Fable, GPT5.6, ...)
Minion (Sonnet 5.5, GPT5.4-mini, ...)

The orchestrator decides what needs to be done
The minion(s) go and execute
The orchestrator checks the result

Even with this setup, coding tasks can rack up to $80/h.

But the work that was done would take an experienced coder maybe 2 weeks.

Crazy world we live in.

[VP Martin @VP_Martin1]

@0xDesigner What could go wrong

[VP Martin @VP_Martin1]

@OpenAI Terra, Luna causing PTSD for crypto bros

[VP Martin @VP_Martin1]

I’ve got the same process as you. Have not found any success or efficiancy with running “loops”.

It did start using Linear recently, connected to Codex/Claude. And in my agents.md something like:

- for every new task get or create Linear issue
- for every major update on the task add comment to the issue
- whenever I say the task is done, set issue to done.

Per project added context in agents.md for the Linear project name.

-> it helps me track my work
-> keeps context somehow centralised because I use Claude/Codex and Opencode interchangingly for maxing limits and free api usages