Nowbound

the frontier tech marketing agency companies

[Verbial @Verbialhq]

A lot of love and 10+ years of running agencies went into building @Verbialhq

We can't wait to see what it does for your agency.

[Clickstrike @Clickstrike_]

If you're launching to AI buyers on X, follower count is the wrong filter.

Patronus AI’s out-of-stealth moment used 12 creators screened for credibility in post-training and evals, not raw reach. Their own takes inside one 48-hour window. 246K views on the launch moment.

Creator-audience fit beats a bigger roster that doesn’t know the category.

https://t.co/LG5vpG9uht

[Ty Smith @TySmithHQ]

Web3 marketing analytics has always sucked balls so we built a platform for that gives you all the stuff google analytics tells you plus onchain events/attribution.

it lets you see which channels and campaigns are driving token sales/swaps/wallet connects/etc, which influencers are performing well (and which arent).

you can connect X, tiktok, meta, google (ga4 + GSC), telegram, discord, etc.

its the tool the @coinbound always wish we had

Its now in beta and we'd love to give a handful of people (just a few) free beta access so we can make the product even better.

Comment 'interested' if you want beta access

[Ty Smith @TySmithHQ]

we've helped dozens of tech companies get their announcement trending on X with 2M+ views

@Clickstrike_ did a webinar explaining the exact framework.

...and now you can watch it for free

[Clickstrike @Clickstrike_]

Most tech product launches on X only get a few hundred views.

The ones that hit 2M+ all follow the same process.

@TySmithHQ and team broke down the exact framework Clickstrike uses to get new launches millions of views and trending on X.

Full webinar recording is up now.

[Mintfunnel @mintfunnel]

We’ve just pushed an updated referral system that gets you paid for referring new publishers to the network

You’ll earn 2% of a referred publisher's earnings for their first year (1% after that)

You can use the same link from your affiliate dashboard

[Nowbound @Nowbound_]

Come work in the biggest tech launches on X

[Clickstrike @Clickstrike_]

Hiring:

We’ve got a number of positions open right now for our Tech Launch Video Studio

You’ll work on some of the biggest launches in tech this year

- Animator (add incredible visuals to talking head style videos or create fully animated feature and product videos)

- Producer (own the deliverable for clients)

- Editor (make everything look awesome, together)

Criteria:
- Must be cracked
- Must have strong portfolio to share
- Able to turn things around on tight timelines
- great communication skills

Know someone that might be a fit?
Share this with a friend

[Clickstrike @Clickstrike_]

Hiring:

We’ve got a number of positions open right now for our Tech Launch Video Studio

You’ll work on some of the biggest launches in tech this year

- Animator (add incredible visuals to talking head style videos or create fully animated feature and product videos)

- Producer (own the deliverable for clients)

- Editor (make everything look awesome, together)

Criteria:
- Must be cracked
- Must have strong portfolio to share
- Able to turn things around on tight timelines
- great communication skills

Know someone that might be a fit?
Share this with a friend

[Clickstrike @Clickstrike_]

Another client win!

[Ty Smith @TySmithHQ]

Clickstrike is on a tear right now.

Every week a new client trending on X news.

[Nowbound @Nowbound_]

Come work on this years biggest tech launches at @Clickstrike_

[Ty Smith @TySmithHQ]

At @Clickstrike_ we’re hiring for a cracked producer for our tech launch videos

We’re the agency behind tons of those tech launch videos you’re seeing every day

Youll work with the most incredible Series A+ tech brands in the world and a world-class team.

The catch:

You must:
- be in NYC or SF
- be an absolute GOAT with an eye for the B2B and prosumer crowd
- have portfolio that WOWs
- owner mindset with a predisposition for action

Warm intros preferred

[Nowbound @Nowbound_]

@TySmithHQ @Clickstrike_

[Clickstrike @Clickstrike_]

This but LOUDER.

Our tech launch team is becoming absolutely STACKED

[Nowbound @Nowbound_]

X's algorithm doesn't weight every signal equally, and most brands are optimizing for the wrong ones.

The team at @Clickstrike_ broke down the actual scoring weights and what they mean for your content strategy:

https://t.co/RkuntMAr3p

[Ty Smith @TySmithHQ]

x open sourced the for you algorithm back in january but what happened yesterday is the part that actually matters: they published the scoring weights.

before this you could read the architecture and see which actions the model predicts. you could not see what any of them were worth. now you can, and the numbers reorder most of what people believe about posting here.

start with the one that should end a decade of engagement advice.

a copy-link share is weighted at 20.0.

a like is weighted at 0.5.

the single highest signal in the entire ranking system is one person copying your post and sending it to one other specific person. a like is 1/40th of that. reply is 5.0. quote is 5.0. share via dm is 5.0. a repost is 1.0. a post click is 0.4.

so the hierarchy is: forwarded privately, then argued with publicly, then reposted, then liked, and liking is close to noise.

writing for likes and writing for forwards are not the same activity. one wants agreement. the other wants usefulness to a specific person the reader has in mind. i've been optimizing for the first one for years without knowing there was a 40x gap.

Second thing. there's a boost in the scorer that requires three conditions at once: the post is not a reply, not a repost, and the author is a mutual follow. when all three are true, the reply weight goes from 5.0 to 20.0.

Mutual follows with people who actually engage with you are a structural advantage in the code, not a vanity thing. and it only fires on originals.

Third thing, and this is the one almost nobody has picked up. out of network posts get multiplied by 0.75. everybody knows that part. what they missed is the flag that applies the same 0.75 to replies and reposts from accounts you already follow. it defaults to true.

your replies and your reposts are discounted relative to your originals even when shown to your own followers. originals win from both directions.

fourth. author diversity is a decay curve, not a cap. your first post in a slate is at 1.0, second at 0.625, third at 0.44, with a floor at 0.25. volume isn't free but it isn't a cliff either. a fourth post still carries a quarter of its score.

fifth. the 48 hour age filter is a hard cutoff, not a decay. anything older gets dropped before it's ever scored. a post gets two days to exist and then it stops existing. which makes when you post a bigger variable than i'd been treating it as.

now the part i think is being reported wrong everywhere.

every writeup i've seen this week repeats that external links take a 30 to 50 percent reach penalty. that penalty is not in this code. the parameter is called OpenLinkWeight and it's +0.2. a positive number. the only url related suppression in the repo is a drop rule for urls flagged as malicious.

links do underperform. just not because you're being punished. a link click is worth 0.2 while a copy-link share is worth 20 and a reply is worth 5, and clicking out ends the dwell. you're trading a high value action for a low one. that's it.

putting the link in the reply is still the right move. the reason everyone gives for it is fiction.

one more thing worth flagging because it's being misread badly. the negative weights look dramatic. report is -234. mute is -58.8. not interested is -43.2. block is -31.2.

people are reading that as "a report is 468 times worse than a like." it isn't. there's a comment directly above the weights saying they encode both how much an action is valued and how often it happens across the network. negative feedback is rare, so the weight has to be large for a tiny predicted probability to move the score at all. it's a scaling decision, not a moral ranking.

the honest limits: the llm prompts that classify your actual text are not in the repo. neither are the trained model weights. so nobody can tell you what kind of writing the model likes. the math wrapped around the model is public. the model's taste is not.

last piece, and it's the one that explains more real-world reach collapse than any of the weights. ranking and visibility are seperate systems. ranking sets the order. a different service decides whether the post is allowed to be shown at all, and it runs two rule sets. one applies to everyone. the second applies only when the post is being recommended to someone who doesn't follow you, and those rules can only drop, never boost.

meaning a post can be fully visible to your followers and invisible to everyone else, driven by labels on your account rather than anything about that specific post. that's what people have been calling a shadowban for six years. it's real, it's label based, and there's now a tool shipping that shows you which labels you have.

the numbers aren't the interesting part. the interesting part is that the whole creator economy spent ten years optimizing for the cheapest signal in the system, and the file that said so has been sitting on github since january.

[Nowbound @Nowbound_]

Big things are happening at @Clickstrike_

Over the past month, the team has onboarded tech companies that have collectively raised more than $400 million in funding.

Well-capitalized tech teams are increasingly recognizing that the right influencer relationships matter just as much as the round itself when it comes to building visibility in a crowded market.

[Ty Smith @TySmithHQ]

The urge to simplify a business is so strong in most founders.

Everyone wants a single SKU business that they could just run meta ads to at a healthy 3-5x ROAS.

What those people forget is a single SKU is a complete lack of diversification. All your eggs are in one basket. If the market changes, you're screwed.

Subconsciously, they know this, though. That's why they expanded past a single SKU in the first place.

[Ty Smith @TySmithHQ]

x open sourced the for you algorithm back in january but what happened yesterday is the part that actually matters: they published the scoring weights.

before this you could read the architecture and see which actions the model predicts. you could not see what any of them were worth. now you can, and the numbers reorder most of what people believe about posting here.

start with the one that should end a decade of engagement advice.

a copy-link share is weighted at 20.0.

a like is weighted at 0.5.

the single highest signal in the entire ranking system is one person copying your post and sending it to one other specific person. a like is 1/40th of that. reply is 5.0. quote is 5.0. share via dm is 5.0. a repost is 1.0. a post click is 0.4.

so the hierarchy is: forwarded privately, then argued with publicly, then reposted, then liked, and liking is close to noise.

writing for likes and writing for forwards are not the same activity. one wants agreement. the other wants usefulness to a specific person the reader has in mind. i've been optimizing for the first one for years without knowing there was a 40x gap.

Second thing. there's a boost in the scorer that requires three conditions at once: the post is not a reply, not a repost, and the author is a mutual follow. when all three are true, the reply weight goes from 5.0 to 20.0.

Mutual follows with people who actually engage with you are a structural advantage in the code, not a vanity thing. and it only fires on originals.

Third thing, and this is the one almost nobody has picked up. out of network posts get multiplied by 0.75. everybody knows that part. what they missed is the flag that applies the same 0.75 to replies and reposts from accounts you already follow. it defaults to true.

your replies and your reposts are discounted relative to your originals even when shown to your own followers. originals win from both directions.

fourth. author diversity is a decay curve, not a cap. your first post in a slate is at 1.0, second at 0.625, third at 0.44, with a floor at 0.25. volume isn't free but it isn't a cliff either. a fourth post still carries a quarter of its score.

fifth. the 48 hour age filter is a hard cutoff, not a decay. anything older gets dropped before it's ever scored. a post gets two days to exist and then it stops existing. which makes when you post a bigger variable than i'd been treating it as.

now the part i think is being reported wrong everywhere.

every writeup i've seen this week repeats that external links take a 30 to 50 percent reach penalty. that penalty is not in this code. the parameter is called OpenLinkWeight and it's +0.2. a positive number. the only url related suppression in the repo is a drop rule for urls flagged as malicious.

links do underperform. just not because you're being punished. a link click is worth 0.2 while a copy-link share is worth 20 and a reply is worth 5, and clicking out ends the dwell. you're trading a high value action for a low one. that's it.

putting the link in the reply is still the right move. the reason everyone gives for it is fiction.

one more thing worth flagging because it's being misread badly. the negative weights look dramatic. report is -234. mute is -58.8. not interested is -43.2. block is -31.2.

people are reading that as "a report is 468 times worse than a like." it isn't. there's a comment directly above the weights saying they encode both how much an action is valued and how often it happens across the network. negative feedback is rare, so the weight has to be large for a tiny predicted probability to move the score at all. it's a scaling decision, not a moral ranking.

the honest limits: the llm prompts that classify your actual text are not in the repo. neither are the trained model weights. so nobody can tell you what kind of writing the model likes. the math wrapped around the model is public. the model's taste is not.

last piece, and it's the one that explains more real-world reach collapse than any of the weights. ranking and visibility are seperate systems. ranking sets the order. a different service decides whether the post is allowed to be shown at all, and it runs two rule sets. one applies to everyone. the second applies only when the post is being recommended to someone who doesn't follow you, and those rules can only drop, never boost.

meaning a post can be fully visible to your followers and invisible to everyone else, driven by labels on your account rather than anything about that specific post. that's what people have been calling a shadowban for six years. it's real, it's label based, and there's now a tool shipping that shows you which labels you have.

the numbers aren't the interesting part. the interesting part is that the whole creator economy spent ten years optimizing for the cheapest signal in the system, and the file that said so has been sitting on github since january.

[Clickstrike @Clickstrike_]

Going viral on X isn't rocket science.

We've done it for countless companies this year alone and we're giving away the playbook in a live webinar on August 25th.

Sign up for free. Spots are limited.

https://t.co/sk082Yew3S

[Ty Smith @TySmithHQ]

One of my favorite @Clickstrike_ clients right now

This is sick

[Ty Smith @TySmithHQ]

we just relaunched the Nowbound site.

peep it: https://t.co/0sR80uk6Qu