Using remote Codex to control Claude, which then controls CapCut...
Editing two 30-minute videos with transitions and audio, all from my mobile phone while walking down the street feels insane.. π
People are losing their voices. Even putting together a sentence and responding as yourself is starting to get handed off to AI.
Derrick.dk ποΈ retweeted
Don't use this
Build your own software using their tokens
So when their tokens become too expensive
You can switch to your own GPUs
Your compute. That you control
So you yourself can't be controlled
You need to control your stack -- or you become enslaved
I'll be honest, Webflow MCP is one of the best products I've ever used. Their API is so clean to work with for scaling sites
Higgsfield is so spammy. The UI feels like a pirated movie website with popups everywhere. Greedy mess.
Derrick.dk ποΈ retweeted
I just found a repo that lets AI agents control a VIRTUAL IPHONE.
Not the iOS Simulator.
Actual virtualized iOS on Apple Silicon.
Screenshots. Touch. Swipes. Typing. Apps.
Thereβs even an MCP server for coding agents.
+633 stars TODAY.
github.com/Lakr233/vphone-clβ¦
'Let go your editor' followed by 'the last 5% still needs your ear' is mad.
Mate, if you're reviewing the cut and fixing the timing, you're the editor now. You just fired the person who was good at it.
let go your $1600/mo video editor
davinci resolve MCP and GPT-6 Astra can now do video editing
here's how to get codex to do it
buy resolve studio 21.1.
file > setup ai assistants. open chatgpt desktop, switch to work mode, tell codex to test the davinci resolve mcp connection
hand it the footage folder
from there it can transcribe the take, put a-roll and screen on separate tracks, cut dead air, pick cutaways, pull stills and b-roll that match 2β3 refs you attach, build titles and lower thirds, key greenscreen, do splitscreen, animate graphics, apply a power grade, mix audio to youtube loudness, and render 4k
a raw 75-minute recording can come back as a finished cut
a folder of clips can come back graded
a few hours of travel footage can come back as a 20-minute video in under 10 minutes
it talks to resolve through mcp
media pool, timeline, color page, render queue and no clicking around the gui
prompts that work:
cut the pauses.
match b-roll to these refs.
use this power grade.
this shot is overexposed, that one has a magenta tint, fix contrast, analyze and balance it again if it still looks off.
captions and titles.
greenscreen or splitscreen, or pick the layout.
mix it. render.
save that as a skill so you are not rewriting the same brief every video.
the last 5% of timing still needs your ear
$295 once for studio. astra tokens after that. studio is required for the official mcp. community servers exist. long runs chew a 20x plan, so reuse the skill
βWrites in your voiceβ has become the new βpowered by AIβ. Everyone says it, almost nobody proves it.
Matching my sentence length is not my voice. Knowing what Iβd call bullshit, what Iβd care about and when Iβd say nothing is.
Most tools learn the wrapping. The judgement is still yours.
Personal AI gets interesting when it stops asking for more screen time.
No feed. No notifications. No new app to check. Just the useful bit, printed while you sleep.
Sometimes the best AI interface is paper.
Instead of checking my phone first thing, I set up an agent to make a "morning newspaper".
It prints while I sleep. I used grokbot @bot to pull from my calendar and email
I made it available if you wanna try it π
Today with AI I booked 2 flights, a hotel, migrated 2 emails and purchased 2 domains.
Now overnight I'm setting up 2 apps to be built and rebranding 2 websites.
What a crazy time to be alive.
Is there a Higgsfield alternative?
I hate the fact that they're everywhere so would rather pay someone else
Uber shuts down two African markets and launches robotaxis in London in the same 24 hours.
Everyone's debating if AI takes jobs eventually.
Uber already answered that. It's just running the transition quietly enough that most people are watching the wrong headline.
60% MCP usage sounds exciting, but it also shows how blind most analytics still are.
A visit no longer has to be a person in a browser. It could be an agent comparing options, pulling data or scraping the whole thing.
Traffic is becoming the easy number. Knowing which agent calls led to a real booking is the useful one.
Hotelist.com has found some type of market fit
Maintaining traffic even if I don't talk about it and also having lots of MCP usage
Actually 60% is MCP usage! Will investigate, I hope it's real usage and not scraping
Derrick.dk ποΈ retweeted
This will change how sports content gets made and delivered.
The method is Gaussian splatting. It takes video from multiple camera positions and rebuilds the whole thing in three dimensions, viewable from any angle.
Instead of polygons (the way games render), it builds the scene from millions of tiny glowing splats of color and depth. Arcturus runs it across time (4D), so a live game becomes a 3D scene, shown from field level, beside a player, or above the play.
Designers who refuse to learn growth are going to fall behind.
Growth teams that treat design like decoration already have.
A landing page needs to catch attention, make the offer clear and get someone to act.
Splitting those jobs across three teams that barely speak to each other makes no sense anymore.














