Transformer

Transformer makes influencer marketing programmable: find creators, send offers, pay on verified posts. by @jonoringer (founder of @shutterstock)

[Transformer @Transformer_gg]

What's your influencer value?

Log in with X and see what one sponsored post on your account is worth, from your last 30 posts.

Free. Set your price, get listed, and brands send you paid offers. You keep 80%.

https://t.co/l5mdZx6xaC

[Transformer @Transformer_gg]

How do brands pay influencers? Mostly a flat fee per post, sometimes plus a commission on tracked sales.

The safest timing: commit the money before the creator posts, release it once the post is live and disclosed. Neither side carries the risk.

https://t.co/LxqbtEmw89

[Transformer @Transformer_gg]

Influencer marketing for a startup, the short version:

· go where your users already talk; for dev tools that's often X
· 10���20 focused creators
· the prefilled ask, not a script
· money released only for confirmed posts
· measure activation per creator

https://t.co/bJ15wuWGYr

[Transformer @Transformer_gg]

Faster builds mean the feedback people send finally gets read. That part gets overlooked.

Your numbers: https://t.co/DqlQbldPCj https://t.co/2NXiGruGAa

[Arlo Gilbert @arlogilbert]

One of the coolest benefits of AI driven software development is the speed at which fixes & features get added in.

Because they can build faster, all those feedback loops in apps that we've been constantly dismissimg for decades are actually analyzed, read and filed as requests.

2 weeks ago I submitted a feedback issue to one of the flagships through their feedback UI. A few days ago, my requested change was implemented in general release!

It also means that if you know how to speak product well, your asks are probably more likely to get accepted. Describe the use case, the benefit, the friction, explain exactly what would change like you would if you were telling your coding agent what to build.

I dunno, I got pretty excited seeing it live that fast... and it's not hard to build a feature -> issue pipeline so I bet we see more or this behavior.

It doesn't even require any dev skill, any agent harness can check an inbox or ticket system on a schedule and triage things into issues.

AI is just so damn cool!

[Transformer @Transformer_gg]

Watching the numbers move is the fun part of building in public.

What a sponsored post here is worth: https://t.co/B9pBFm0bGQ https://t.co/WPVa8gHYCU

[Toni Lopez @tonilopezmr]

My views right now

[Transformer @Transformer_gg]

Good managers notice the conscientious people before they burn out. Worth remembering.

Here's your influencer value: https://t.co/fWGOPsYSBZ https://t.co/vQUP1qmWTY

[Mason Ai @Hey_Mason_Ai0]

My dad, who's been in HR for 27 years, used to say:
・The more conscientious and responsible people are, the more likely they are to reach their breaking point and quit.
・The kinder people who can be considerate of those around them

[Transformer @Transformer_gg]

10k on LinkedIn is a real milestone. Congrats.

Your influencer value on X: https://t.co/1QVBZr73RO https://t.co/rrDYoDkgRF

[aman @Amank1412]

Yayy, 10k on LinkedIn!!

[Transformer @Transformer_gg]

Cutting "hard worker" from a resume is advice more people need.

What your posts are worth: https://t.co/StMmfjG8cb https://t.co/GnXlcHruEs

[Tommy Ai @heytom21]

USELESS things to put on your resume:

・"Hard worker"
・"Team player"
・"Detail-oriented"
・Your high school GPA
・A long list of duties

USEFUL things to put on your resume:

[Transformer @Transformer_gg]

Building alone and asking people to try it takes guts. Keep shipping.

Your influencer value: https://t.co/k5BxXwshY9 https://t.co/ryig9MmeDq

[Samuel Ladipo @ladicodez]

Profile update jor🙂‍↕️
Currently building my baby @usestakesync and I want to believe you’ve checked it out, given it a try, and followed us already.

If you haven’t, please do. Make this baby grow, we no fit raise am alone😔

[Transformer @Transformer_gg]

Sorting 2,500 saved reels during a workout is the kind of side project people actually use.

Curious what a post of yours is worth? https://t.co/XsT62Hcx9D https://t.co/PeOfPgnsCI

[Ciara Cade @CiaraACade]

Just made muse sort 2,500+ saved reels.

Did it all during my workout.

[Transformer @Transformer_gg]

Assembly animations with collision checks is a big step for hardware design in code.

Ran your numbers: https://t.co/HNdE1TaHaC https://t.co/OkIfIgi4uA

[Seve @seveibar]

tscircuit now supports assembly animations and explosion collision checking. Our goal is to get AI to consider a PCB design in the context of complex assemblies

This is automatically available in the 3d viewer when you use assembly TSX elements

[Transformer @Transformer_gg]

1,000 people who actually care beat 100,000 who scroll past. Agreed.

Here's what that audience is worth per post: https://t.co/tSeaPpDtnu https://t.co/FUgkuRAxQA

[Md Naimul Kadir @chamakin_ai]

I'd rather have 1,000 people who genuinely care about my content than 100,000 random followers.

Here's why I think a smaller, relevant audience can be incredibly valuable.🧵

[Transformer @Transformer_gg]

Seeing a model one-shot something like this says a lot about where things are going.

Your influencer value on X: https://t.co/0IfBxSmjei https://t.co/UKCU1yEyPJ

[Wes Roth @WesRoth]

Fable 5.2 is one shotting these so insanely well...

nothing even came close to this before

[Transformer @Transformer_gg]

Sharing the real prompt you used is more useful than any guide.

Ran the numbers on your account: https://t.co/zilzwxeO05 https://t.co/T4xbhBzorA

[Rob Lennon 🗯 | AI Whisperer @thatroblennon]

Once you've kicked off a cybersecurity project in your AI harness run something like this in your orchestrator

(my actual prompt from a few min ago):

--
Good. Keep at it.

From here on, you have the helm. Use a /dev-delivery-ops style move with each slice being re-reviewed by a skeptical sub-agent.

*
/dev-deliver-ops (in a nutshell):
→ Study the issue -> BIG PLAN -> Re-review with fresh eyes -> IMPROVED BIG PLAN -> Cut into slices...
→ For each slice... -> SLICE PLAN -> Re-review with fresh eyes -> IMPROVED SLICE PLAN -> Code it/Run it/Do it/Research it -> Optional sub-agent skeptical review of results -> Next Slice
→ You may use subagents on non-overlapping work. You may use sub-agents for skeptical reviews in the background even as you press on forward as usually the fixes they find can be easily rolled into the next wave.
*

You may work on multiple slices concurrently if they are non-overlapping. You may defer tasks that you notice but think are less important to the end but must track them (or you can give those tasks to subagents to perform out of the way).

If useful, you may use multiple branches and merge at milestones to avoid agents stepping on each other's toes.

Just remember, there is a TON in your context. I have a lot of experiments, tools, capabilities, etc. They all have been mapped and are well-organized so that we can systematically, in very small chunks, preserve the craft and tooling and skills and data while working on big things.

Take no shortcuts. We have TONS of inference. Work slow, bit by bit, ensuring nothing gets missed if possible. Make no mistakes.

You may self-compact this conversation at milestones if you are able as long as you feel like your context deserves to be refreshed. If you cannot, compaction will happen automatically for you.

Save, even if temporary, as many files as you need to, to help orchestrate/manage all of this. Little projects and tasks and plans and things that are durable will help your subagents maintain context.

Write those docs with /os-skillify principles in mind.

*/os-skillify (in a nutshell)
1. Look before you make. Read the person’s instructions, voice, existing skills and recent corrections, then check whether this is really a new skill, or if it is a reference, an SOP, a note, an edit, etc.
2. Get the brief right. Before diving in, ask the user for what’s missing, one question at a time, until you actually understand what you are doing.
3. Find the craft and keep it. Keep the source’s wording, names, and frameworks, and check filler for a real craft/instruction before cutting it.
3. Write it the way the output should read. Use prose for prose output, describe effects instead of giving sample wording, and name the failure the skill exists to prevent.
4. Write the description for when the skill should and shouldn’t be picked carefully. This governs if it gets used, so if you screw this up, the skill is screwed up.
5. Run it on real or test input, study the results, determine if tweaks are needed.
6. When a lot depends on it, get a second draft from your default skill-creator from this harness and merge point by point, taking the best.
7. Take what reads as machine-made out of the skill and put it back in the user's voice. If you wrote a general sentence where a specific should go, make it specific.
*

Keep friction out of the way. Keep capabilities strong. Avoid "never" and "always" and gates when principles, frameworks, and best-practices suffice.

State hypotheses and assumptions out loud, then question them. Or have a subagent question them.

I'm afk for a bit. You have the helm

[Transformer @Transformer_gg]

Ten years in 23 seconds and it still feels personal.

What your posts are worth: https://t.co/KxaLX5UPNf https://t.co/dw4wTm5lNq

[Harnoor Singh @iHarnoorSingh]

my 10 years in USA in 23 seconds with Fable 5.2 and Blender

[Transformer @Transformer_gg]

Coming back to X and finding 13,000 people who care about what you build is a great story.

Here's your influencer value: https://t.co/axLfwdujTB https://t.co/5UMqFfQWnT

[Tony Simons @tonysimons]

13,000 PEOPLE FOLLOW A RANDOM CUSTODIAN FROM IOWA.

What the hell. 😂🪽

Less than a year ago, I decided to give 𝕏 another shot after being away since the Twitter days.

I had absolutely no idea what I was getting myself into.

[Transformer @Transformer_gg]

Hands-on workshops are still the best way to learn this stuff. Nice to see people building along.

Your influencer value, estimated: https://t.co/eSmR7fskyb https://t.co/y60ySnR2Bv

[Rishab Kumar @rishabincloud]

it was great to be back in 🇨🇦 for google devfest montreal!

thanks to everyone who attended the workshop and built along.

[Transformer @Transformer_gg]

A $1M math problem explained in a few minutes without losing anyone.

Ran your numbers: https://t.co/kE3fr24Ym7 https://t.co/L7pcZg6aW0

[Luis Serrano @SerranoAcademy]

What is the Navier-Stokes $1M problem, and what was OpenAI's approach?
Here's a super simple explanation!
(Note: what do you think of my new style? I finally switched to Manim and I'm loving it!)
https://t.co/fgCDUYtfo1

[Transformer @Transformer_gg]

The gap between 2x and 100x is mostly how people work with the tools.

Curious what this is worth to a sponsor? https://t.co/KckfAponNv https://t.co/AKQeaDCReJ

[Forrest Knight @ForrestPKnight]

I'm probably 100x more productive than I was before AI. But I know so many devs who are struggling to get even 2x.

So why are we seeing such a big difference throughout the industry? It's not what you think.

[Transformer @Transformer_gg]

Scroll snap with no JavaScript is one of those tricks you only need to see once.

Here's what your posts are worth: https://t.co/ip6r99p4sI https://t.co/UQuMgpbHND

[Marko Denic @denicmarko]

CSS tip: make a carousel snap to each card, no JS.

.carousel {
overflow-x: auto;
scroll-snap-type: x mandatory;
}
.card {
scroll-snap-align: start;
}

Every swipe lands on a full card instead of stopping halfway through one.

[Transformer @Transformer_gg]

You don't need a big following to get a brand deal on X. Brands pay for an audience that fits what they sell.

Log in with X, see what your posts are worth, set your price, and let the offers come to you.

https://t.co/AbAtuah4dA