CEO at Slite (YC18) | We build the first knowledge base on autopilot, in remote since 2016!

I'm the founder of Slite & Super (YCW18) 👉 Super.work 

Super.work turn all your company data into answers & AI agents for your whole team.

It comes with the recipes and tools for your team to get instant benefit from AI at work

Slite is your team dead simple yet crazy powerful knowledge base, since 2017 🔥

I'm fortunate to lead a diverse team of exceptional individuals spread across 10+ countries 🇮🇳 🇬🇷🇱🇹 🇦🇷 🇬🇧 🇳🇱 🇫🇷 🇳🇴 🇺🇸 🇪🇸 🇩🇪 🇨🇦 🇵🇹 . Our journey includes going through YC W18 batch, raising $15M+ in funding from top-tier investors like Spark Capital and Index Ventures, and serving tens of thousands of teams monthly.

Christophe Pasquier

$46M valuation at last raise.
$15M invested.
$4M+ ARR with 50% YoY growth and 115% NRR.
At these numbers, we could exit for a 5-12x multiple depending on strategic interest.

In any scenario, I'd walk away with a meaningful payout.
So why don't I?

Every time I think it through, I ask: what would I actually be trading?

1/ I get to pick and grow an - incredible - team
Most exits come with earn-outs and culture shifts that grind people down. I've seen it happen countless times with friends.

Our team is exceptional, it's a luxury I have to maintain this group of people around me happy and deliver incredible work. I'm not willing to trade that for a payout if we all land in the wrong place and lose this.

2/ We have the chance to push our category to new frontiers
We're years ahead in knowledge management and enterprise search, shipping things no one else has.

We did it in the last years, and will do it again in 2026.

Acquisitions rarely protect that kind of momentum and innovation is usually the first thing to die. And if that was the case, it would be the worst place to work for me.

3/ We're in a lucky position
We're profitable, extremely ambitious, and all this with low pressure. That's a very rare place to be in. Trading independence for someone else's roadmap doesn't make sense when you're already on solid ground.

This is one of the hardest question to navigate as a founder, I would never have written this even 2y ago, and I might be in a different place in a few years. Today I feel blessed to think this way.

Curious to hear about fellow founders, do you have an "exit strategy" or do you feel the only way to succeed as a founder is to be 100% commited?

PS - If you're a founder and thinking of this topic, one truly great book to check is "Lost & Founder" by Rand Fishkin

CEO Slite & Super (W18) | Most accurate search & workflows on team data

We caught 3 permission bugs this weekend that staging would never find.
My CTO wanted to build the system that made this possible for years, now I get why.

When you're refactoring permissions logic on a mature codebase, you can't test every edge case. Staging doesn't have realistic org structures. Unit tests can't cover thousands of users with nested team hierarchies.

Here's what we're running at Slite.

When we want to ship critical changes, we run two versions of our code in parallel.

The proven production system handles the actual response. New system processes the same request silently on the same inputs, testing changes, logging the results for us to see if we missed anything.

Every permission check hits both:
- Same input (user, resource, action)
- Old system returns what the user sees
- New system computes what it would return
- Compare outputs, log mismatches

This weekend it caught 3 critical issues this way, that our developer environment would never surface because test data isn't always able to cover all the edge cases of massive orgs.

My CTO dreamed about this for years. My inner nerd is so happy we're getting there.

I spent 7+ years forcing PLG on Slite.
Super has been SLG from Day 1.
Thing is - there was only one right choice from the beginning. I didn't know how to make the call.

Everyone treats PLG vs SLG like a philosophy. It's not.
Here's how I think about it now:

1/ The viral loop test
Ask one question: Will each free user naturally bring 10 more?
Not through referral programs. Not through incentives. Organically. Because the product spreads by being used.
Lovable has this. Figma had this. Slack had this.
If you don't have a natural loop, PLG becomes a leaky bucket you're constantly filling.
And that's 90% of B2B companies.

2/ The contract math
For SLG, the test is simpler:
Can you sell yearly contracts starting around $15k+?
If yes, the 1:1 time is worth it. If no, you can't afford the sales motion anyway.
PS - the number depends on your retention and LTV, but the principle remains.

3/ The setup question
This one matters more in the AI era:
Does your product require dedicated time helping customers configure it?
If setup is complex and ongoing, you need the price point that justifies it.
And if that setup isn't temporary (something you're fixing with better product), then you're sales-led whether you like it or not.

4/ It's been decided for you, already.
But stop treating this like a strategy decision.
Look at your product. Look at your price. Look at your loops.
The motion is already there. Your job is to recognize it.
Super was SLG from Day 1 because the product demanded it.

Not smarter. Just listening.

Our Super homepage looked slick in dark mode. Conversion dropped.


I'm in love with the brand we developed for Super. Dark backgrounds, beautiful gradients, space-themed colors, the whole nine yards. Our design team crushed it.

But we developed it before our May public launch. Since then our # of users exploded and we saw the best ones - by far - being non developers.  

They're ops managers, support leads, sales teams - normal people here to get instant answers and accelerate the rest of their work. 
And we attract more and more of these persona too!

Problem is, we see our demos generating terrific impressions but our homepage conversion rate dropping. 
We looked at this and emitted the hypothesis that our ICP might not want a terminal aesthetic. And that maybe, the whole vibe of Super.work felt like not thought for them.

So let's see! We ship our new homepage (the right one) next week, with this and a simplification of our value prop. 

We'll see what it gives, and place your bet, will it work? Is this worth challenging the whole visual brand?

Last year - I made a resolution
Share what feels uncomfortable.

We only see wins and launches and funding rounds on Linkedin.
Never the pivots, the hard calls, the tough bets.

So I decided to share just that - all the real things behind building.
And last month, you all were incredibly kind to me.

I got so many messages about how my posts resonated with people.
And what made it special was

Many of the messages were from founders and builders I deeply respect.

Experienced founders who'd done millions in ARR and share similar challenges, some we rarely see shared in the open.

New founders who found it comforting to know that mistakes are a big part of building.

I realised that the internet has been making all the real builders feel inadequate when building is messy, slow, and HARD - all at the same time.

In 2026,
It's my resolution to double down on the honesty
No corporate speak, no inflated metrics
I'll share things I do everyday, and the things I learnt after years of expensive mistakes.

And in case you don't know
I'm Chris, co-founder at Slite (a beautiful Knowledge Base) and Super.work (best AI search on all your team's tools).

We're a 20 person remote team, profitable, globally-distributed, and building 2 products we're really proud of.

I'm gonna publicly share the real stuff even more this year.
Follow if you'd like to come along for the journey.

We throw dinners with top tier operators every month.
Last week was our 3rd edition with Chiefs of Staff in Berlin, here are the nuggets.

→ What is a Chief of Staff?
If you're not familiar with this, quick catch up: they're one of the most interesting roles emerging in tech startups and scaleups in the last 5 years. 
Usually a Chief of staff is attached to one of the C level and acts as their right wing: they'll work on their project, with their authority, but with much more time to properly execute. But they - for the large majority - don't have people to manage.

I absolutely love this role: think of it like some of the smartest folks in the company, with the authority of the CEO, here to solve the hardest cross functional problems hurting teams every day but owned by no one. 

→ When CoS act as silo-breaker
One of the best story of this evening, one of our guest was Chief of Staff to the CTO of a 400 ppl company, and it happens that each of the 5 C level has one.
Result is they actually spend time between them to detect and course correct the issue of their leadership team.
With the right C level and chief of staff this is an incredible super power for any scaling team.

→ But the downside of this..
For other CoS, this necessity to manage their leadership team is very hard and can be a curse more than a blessing.
Projects are shared by C levels in cryptic manners and conflicts between C level has been raised by some as one of the toughest challenge they have on their plate.

→ Top challenge for 2026 remains internal communication
We heard this in the 2 previous dinners, and this is still one of the top problems to be solved for this cross operational role: 
→ accelerate onboarding
→ remove the noise of repetitive questions
→ avoid the unnecessary friction between departments.
(not really a surprise as we deploy super.work through a lot of chief of staff and this is one of the reasons to learn more from them)

I loved this format, if you're a founder, throw these dinners with your top buyers, make them have a good time and learn from them!

It was a blast to meet you all Henriette Brune Jan Philipp Kuhnert Trang Thuy (Rachel) Dang Joon Knapen Mitja Thomas Tabea Severin Anna Jopp Eva Sayre Fadeelah Al-horaibi

Hope you enjoyed it! (and if I missed a key insights you got from this, share in comments ;) )

I have 2 AEs closing deals all day. 
I banned the team from interrupting them. 
Gave them this instead 👇

Everyone needs something from Sales.
GTM needs attribution. Product wants early users feedback. CX needs customer expectations. Execs needs results reports.

So I built a workflow that automatically 👇
→ Goes through Attio, sales recordings, notes, Slack
→ Looks for signals for each department
→ Creates a one-pager with 2 main things each department should know from last 2 weeks of sales conversations.

Here are 4 most interesting things it gives

1/ For marketing
The workflow pulls from last week's Attio data, Slack mentions, and call transcripts.
Then generates:
Which channels drove conversations
How prospects actually describe their problem
What messaging made them reach out
No more guessing which campaign worked ✅

2/ For product
Every feature request. Every "I wish it could..." moment.
Surfaced from real conversations. Ranked by frequency.
You'll hear patterns you'd never catch manually. The roadmap starts writing itself.

3/ For customer support and success
What objections keep coming up. What's confusing prospects before they become customers. What expectations are being set that CS will inherit.
Fix it in sales, save hours in support 😅

4/ For dev
Technical blockers and integration requests.
Every "does it work with..." question. Every security concern. Every API ask.
No more secondhand requirements through three Slack threads.
The output is simple. A bi-weekly digest with:
→ Top requests
→ Attribution insights
→ Excitement signals
→ Deal blockers

Each department gets points. Relevant to them. Actionable.
All from one automated workflow.

If you want this for your team, DM me.

Consensus killed one of my products.
Here are 2 stories, 1 fail and 1 win, 2 years apart.
1/ 2022: Failure by compromise
We went to build a major feature of Slite, Discussions. A way to bring the key decision-making discussions back in your documentation.

Half the team wanted to build them as an extension of docs. The other half wanted it as an extension of comments.

We settled somewhere in the middle.

The result: a Frankenstein. Hours of endless debate. ICs unable to move because nothing was clear.

Consensus killed our product for 1+ year.

2/ 2024: Win by committing.
I pitched starting a new product on top of our existing one.

Half the team thought I'd lost my mind.

We were struggling to unlock growth for Slite. My solution? Build a second product that would take 14 months and split our focus.

The hypothesis - which proved right - was that our buyers would need both. But the objections were valid. "We should double down, not diversify." Multiple heated discussions. Smart people on both sides, all caring deeply about the company.

We discussed, heard all opinions, and took the decision. Not half way.

Similarly when we decided to build it a new debate emerged: should it be inside Slite, or as a standalone?

We picked one direction, and avoided the 2022 pitfall.

The result: a highway for developers and designers instead of a maze, and an incredible product, Super, a year later.


Strong decisions are gifts for your team.

Yes, you can be wrong. But it's much easier to go 100% in one direction, realize your mistake, and turn around than to walk multiple paths simultaneously with low conviction.

Strong decisions let us reach a lovable product, validate demand, and generate revenue at the same time.

That's my ideal is for our team: generate strong decisions, loosely held. Don't be blind, be opinionated.

I made 2 mistakes setting my company goals: setting them on vibes, and basing them on hope.
Now we work differently.

This is my full journey of setting goals - the 2 times when we did it wrong 😅 and how we do it now.

1/ Before traction, we set goals on vibes
Right before our YC batch, we had nothing - no funnel, no distribution, just a launched product. I asked our partner what our demo day goal should be.

Their answer:
"You want to be the Slack for notes, so aim for the sky. Go from 1,5k to 100k users in 3 months."
(not representative of my overall YC experience btw, this is the only bad advice I can recall, and it was probably more due to me as a young founder interpreting this from our convo)

I tuned it down to 30k, but it still put enormous pressure on the team. We just had started, the 1,5k users where volatile signups from our initial PH...
We shipped like crazy and hit 13% week-over-week growth. By any measure, it was the best stretch the company had ever had. But it felt like failure because we missed the number.

I remember feeling frustration instead of celebration, and I know the team felt the same.

That's when I knew something was broken about how we set goals.

2/ After traction, we set goals on hope
At $1-2M ARR, we were powering through and I needed something to energise myself, the board, and the team.

So we said "let's double the business" without a clear set of actions to achieve it.

The issue there is it made us lose focus on foundations, on the quality we needed to keep pouring in our customers and core feature to build our compounding effects.

As a result we added years of legacy to pay back by trying to aim for only very large bets who could radically raise our trajectory.

3/ After traction: goals from math
Now, we start with baseline
→ What stays the same if we do nothing different
→ Existing revenue trends, current retention, predictable growth.
Then we look at specific levers that could change things:
→ Repositioning Slite as a bundle
→ Outbound, SEO, and LinkedIn efforts
→ Having developers work directly on sales deals

Each lever has an estimated impact. The goal becomes baseline plus realistic lever outcomes, not a number we hope sounds impressive.

The past two years, we've hit our goal, last year we did +50% in revenue with great metrics on all sides.

Not because we sandbagged, but because we were realistic about what was achievable. And it cascaded down in improving all our business fundamentals, our key customers, our NRR, our brand positioning, our product stability, our developer velocity etc...

The way I think about it now:
Goals should be just out of reach enough that hitting them feels great
But missing by 5-10% still feels like a win.

If hitting your goal brings relief instead of joy, the goal was wrong.

OpenAI had the first mover's advantage in AI but their last big thing was... announcing ads. Anthopric? Claude Code.
Why is Anthropic - the second mover - winning? When did OpenAI begin losing?

Here's what I'm seeing:

1/ First movers set the map, second movers choose the road
OpenAI defined the category. But they also made every mistake first.
Anthropic watched. Learned. Then picked their lane.
Model quality for coding. Enterprise reliability. Developer tools.
They didn't try to win everywhere. They just won where it mattered.

2/ Consumer success can become a distraction
OpenAI built ChatGPT into a cultural phenomenon. 100M users. Household name.
But that success pulled them into a war with Google and Meta. Massive companies with infinite resources and consumer DNA.
Meanwhile, Anthropic stepped away from the crowd.
They bet on developers and enterprise work.

3/ They shipped protocols, not just products
MCP. Defining papers on RAG. Skills.
They shipped foundations other companies build on.
When you set the standard, you don't chase the market. The market comes to you.

4/ The cult following became a movement
It started with developers who loved Claude for coding.
Then Claude Code dropped. The first true agentic AI for technical work.
Then Claude Cowork. The same power, for everyone else.
Enterprise usage? Already led by Anthropic.

Not because they were first. Because they were focused.
Anthropic kept watching what's missing while OpenAI tries to win a losing battle on 10 fronts.

What if you could use Nano Banana right inside your docs?
Hear me out, you can generate any visualization you need - blog headers, product specs, social media assets, whatever.

I'm probably more excited about this than I should be but it has been a lot of fun 😅

Slite is becoming absurdly powerful for doc writing workflows.

You can already fill your docs with information from all your tools using Super.
You can drop rough notes and auto-format them at once,
You can proofread these docs against the rest of your data,

And now we play with generating images based on whatever you're writing, without switching a single tab.

Here's what this unlocks:
→ Marketing can generate infographics on the fly for their posts
→ Yesterday I read a complex RFC translated in a human readable diagram
→ Devs can quickly visualize a new spec's architecture without drawing rectangles in Excalidraw.
→ Anyone can create a doc cover in seconds.

No context switching.
No opening Gemini in another tab.
Just describe what you need and keep writing.

Shipping soon.
DM me if you want early access 😉

Trying out Nano Banana in Slite

Documentation won't be read by humans anymore
Check this chart. Our team asks exponentially more and writes exponentially less.

In a world of AI, Context is king and you need verified company context.
But manual documentation is dying.

It always was a painful process. You hire smart people to do high-leverage work, and then ask them to spend hours writing docs that no one can find the next day. Nobody wins.

So what if we had best of both worlds, what if the knowledge base maintained itself?
Imagine your Knowledge Base
- automatically detecting that the onboarding guide needs to be updated because of a Slack convo and just updates the doc itself.
- understanding your Granola transcripts and automatically updating key client info in the pipeline tracking docs.

All this to deliver spot-on answers to your team. This idea really excites me.

It's the truest form of what I set out to do with Slite since the day it began.

Have a Knowledge Base that's not outdated or hated - while still being a single source of truth for your entire company.

And now, since the tech's caught up, 
All Knowledge Bases will just... do the writing, editing, updating, of your everyday work.

Here's our 11 step-roadmap to build the first knowledge base on autopilot
0. Make a Knowledge Base so simple everyone documents and reads ✅
1. Build a nested doc system for all types of docs ✅
2. Let teams verify docs - via API/app - recurrently so it stays up-to-date ✅
3. Let teams talk to their docs like they'd talk to a coworker ✅
4. Flag gaps from question asked ✅
5. Go beyond the Knowledge Base, let teams ask about ANYTHING from ANY tool (Super.work) ✅
6. Cross-check docs against real-time data sources - Slack chats, PRs, CRM records ✅
7. Generate suggestion of changes ready to be applied 🔨
8. Detect outdated articles and generate suggestions on autopilot. 🔨
9. Generate suggestions to fill gaps, on autopilot. 🔨
10. Build the first API endpoint so agents can update/edit your docs 🔨

We're at step 7. That's the thesis. 

Do you buy it?

We've all heard that 95% of AI pilots fail.
No one talks about how to be in the 5%.
We've done 100+ AI handholding pilots for Super with great companies this year and of the ones that fail, always boil down to 1 of these 4 points.

1/ The technology sold simply can't solve your use case → impossible for it to work
The AI fundamentally can't do what you need.
Typical case: You try a custom GPT to "find the answer in your docs". If your service doesn't offer semantic search like Super does, it simply can't work.
It's not a training issue or a prompt issue. The architecture can't support the job.

2/ Missing integrations/tools → impossible to deploy
It can't connect to your CRM, can't pull from your knowledge base, can't push to Slack. Your team would need to copy-paste between six different tools to make it useful. Nobody's doing that. The AI stays in the sandbox while everyone goes back to their old process.

3/ Threshold of reliability is not met → impossible to build trust
The AI works in a demo. It falls apart in your actual workflow.
If the AI is right <90% of the times, it means your team has to verify every single output. Which means they're doing their original job plus babysitting the AI. The cognitive overhead is higher than just doing it themselves. After two weeks of spot-checking, they stop using it entirely. Trust never forms.

4/ No rails set by the tool → impossible to adopt
If you say your tool can do everything, teammates won't know what to use it for.
If you say your tool does X better than anyone in the market, they'll know exactly when to log in.
The blank input box is where adoption goes to die. This is why we obsess over step 4.

Super.work's an AI search on company data with a - you guessed it - blank box at boot up. So to boost adoption, we think of core use cases we want to solve and integrate as deeply as we can there.

If you use our web extension, your support will get Super directly injected in their Intercom and Zendesk for instance. Within a few months, this becam eone of the most used triggers by Super customers.

They could've done it before the button too, the button just makes it obvious.
One click, one job, zero ambiguity.

And we replicate this over and over.

I'm still documenting learnings as the space evolves.
If you're in the same boat - I'd love to know your learnings/takes!

Anthropic is the 🐐.
They will turn my mother into a developer.

First they redefined development with Claude code, and now they want all of us to act as this new breed of developers.

If you code you've seen the pattern they cracked: instead of pouring things in context, just a series of simple atomic tools, called sequentially, in a very fast manner. Grep (search), create, read, write. That's it.

Last month, they put this pattern in our Chrome browser. Claude for Chrome can read, browse, load a URL, screenshot a page.

Now they release Claude Cowork with the same core pattern: it can do all the actions you do with your finder, and simple file editing, on top of advanced scripting.

For the everyday user, they get a true feeling of what 'agentic' means for the first time in their life. Ask it to organise your folders, prepare a spreadsheet based on scattered data, generate a summary from your notes, find things locally.

They managed to match the experience that made them kings in the dev space, except for 1 exception.

Claude Cowork needs knowledge.

Claude Code had this knowledge in the form of your entire codebase.
Your code base is clean, cohesive, structured. When you write `functionName`, you can't make a typo, or refer to it by `maFonction`. 

What made the series of search and read pattern work for Claude code is simply not present for Claude Cowork. 
You have no graph of knowledge, your data - and your team's - is a messy pile of docs, databases, and Slack threads.

And that's exactly why we've been focusing all our efforts in the past 3 years to build the most accurate source of knowledge for teams with Slite, and the ultimate knowledge retriever with Super.work.
This takes a lot of effort, you need
→ connections to all your toolstack
→ embedding of your core, unstructured knowledge
→ sql queries for your structured knowledge
→ smart orchestrated search to blend both (if you say what have I been doing you want the agent to know you're an ops leader and find the info int he right channels and tools)
→ verified knowledge (ideally continuously maintained with signals from your toolstack)

This being said I could not be more excited by this release. 
This will be banger - especially with a true enterprise search MCP.

 Thanks for pushing the world of work forward Felix Rieseberg Boris Cherny, and for pushing this whole project so fast!

PRs merged per contributor went from ~30/quarter to ~70/quarter.
Same team. Same headcount. More than 2x the output.
That's when my thinking flipped. How is this so cheap?

$100/month per person to double engineering velocity. That's less than what we spend on coffee.

1/ The math isn't "AI cost vs no AI cost"
It's AI cost vs the alternative: hiring more people, longer timelines, slower iterations.
One senior engineer costs $150k+/year. We're getting equivalent output gains for $24k across the whole team.

2/ The gains compound
Faster shipping mean faster feedback loops. Faster feedback means better product decisions. Better decisions mean less rework.
The 2x you see in the chart understates the real impact.

3/ The tools that actually moved the needle
We tried everything, but in the end we have 2 winners. Claude Code, and Super.
Our dev team is a huge fan and use it in tandem with Super MCP to get company context about bugs and more context right in their coding workflow. No opening Slack.

It saves time, it makes everyone more happy that they get to do more of the work they deem important for them to cognitively engage with.

$24k would feel like a lot for any other tool.
But for the impact, it's the best money we spent all year.

Writer

For fun, I love reading about Creativity, Advertising, AI, startups, and about everything I can get my hands on.

If you'd like to say hi and argue with me about one of my blogs, 
Here's my calendar: https://calendly.com/gupta-ishaan/30min

Ishaan G.

A few too many moons ago, I used to work in an agency. One of my everyday tasks was to be on the lookout for a moment and jump on the bandwagon ASAP for our client. I’d see a trend, maybe whip something up in an hour. Then it’d go through 5 people, 5 revisions, and 4 tone-deaf dilutions. By the time the tweet would go live, the moment would be almost always be dead.

Culturally viral moments happen fast, die faster, and as you fight for attention - you need to prove why you deserve it.

In the same pursuit, some incredibly creative folks over the world went as far as to backspace the draft and vibe-code full apps instead.

If you're a brand, this is your new 'in'. Your witty tweet will be a flop, but your functional product might even get more virality than the original. 

If you're in social, this is your new 'in'. Everyone can write a clever caption. Not everyone can turn a trending topic into a product that gets covered by The Telegraph. 

If you're in SEO, remember that search follows social. It always does. While everyone's tweeting about the moment, you could be the first to rank for the search term that didn't exist 48 hours ago. 

The window between "this is an idea" and "this is an MVP" is shrinking. 

Marketing teams either figure out vibe-coding or start begging for dedicated dev bandwidth if they haven't already.

Ishaan Gupta

Idea to shipping velocity is insane right now

Controversy Marketing has existed longer than the internet, so why can't tech figure it out?
Vince McMahon perfected it in wrestling in the late 80s and 90s. The Montreal Screwjob in 1997 turned him into the most hated villain on television. He didn't fight it. He leaned in. Created "Mr. McMahon" - the evil boss character.

Hip-hop figured it out around the same time. Tupac vs Biggie positioned entire record labels against each other. The corporate suits got rich. Two artists ended up dead.

So when AI became the most divisive topic of the 2020s, it was only a matter of time before someone figured out how to turn that tension into marketing.

Enter Artisan AI.
Late 2024. Billboards across San Francisco: "Stop Hiring Humans." The founder got death threats. Billboards got vandalized. He didn't back down. The company raised $25 million. Hit $6 million ARR. The playbook was validated.

Then came Cluely.
Roy Lee got suspended from Columbia for building an AI tool to cheat on coding interviews. He filmed himself using it during his Amazon recruitment process and posted it on LinkedIn. His philosophy, "If I say extremely crazy shit online, it will make more people interested in me and the company." He hired 700 "clippers" paid per view to repost content across TikTok, YouTube, and Instagram. He called reputation "a thing of the past." Attention was the only currency.

Then came Chad IDE.
November 2025. Y Combinator announces "Chad: The Brainrot IDE" - an AI code editor where you can gamble, watch TikToks, and swipe on dating apps while the AI writes your code. People thought it was a joke. It wasn't.

This is when sh*t hit the fan. Jordi Hays wrote a post titled "Rage Baiting is for Losers."
Cluely was an app for cheating. Chad IDE's only differentiation was gambling. The rage bait was the product itself. 
"Running a successful VC-backed company requires you to build a coalition of people that want to see you win," he wrote. "Rage baiting is the most effective way to get people to actively pray for your downfall."

But McMahon made it work. Death Row made it work.

Why can't tech?

Because wrestling and hip-hop are entertainment. Heat is the product. You tune in to see the villain get what's coming to him. You buy the album to hear the diss track. The controversy is the thing you're paying for.

SaaS is different. You're asking people to trust you with their data, their workflows, their money. You need employees who want to work there. It's hard to build that coalition when half the internet is waiting for you to fail. Rage bait works when attention is the end goal. It breaks when trust is.

I've been off Instagram for six months just found out about Claude's Instagram strategy from a tweet.
If you're a marketing team with the kind of organic hype Claude has on Twitter, you could reasonably think you're done. The "Claude Code is insane" posts are writing themselves. Why spend another dollar?

But they're going after Instagram too. 

Anthropic does great EGC on X. Boris Cherny, Felix Rieseberg, and the team are shipping updates and engaging with AI discourse almost to the point of shaping it. They're building in public and can be found in the comments section of any tweet doing more than 1 million views. SF tech bros know these names by heart and care about shipping speed and seeing quick TAT on a product as big as Claude. 

But they're doing even better UGC on Instagram. They're partnering so much with influencers who can code, building aesthetic little apps, museum diary tools, "building it for myself" content. The vibe is joy-centric and 100% Instagram-core. "Look at this cute thing I made." 246K views on a "POV: your crush just said he overcooked his steak" video about Claude. 

The interesting part is how precisely they've adapted for each platform. They understood that Twitter wants discourse and shipping and technical credibility. Instagram wants vibes and joy and personal projects. And they're executing both at the same time, at a level where each feels native to its platform.

Most marketing teams with this much hype on one platform would ride on it.
Claude's going after everyone.

“We need a Reddit SEO strategy.”

‘Why?’

“AI search engines cite it constantly.”

‘Why do they cite it?’

“Because the answers are actually useful.”

‘What makes them useful?’

“Real people have real conversations there. Anyway, we should figure out how to get our brand in there. Though their mods will probably remove anything promotional.”

‘Right. And if the mods didn’t remove promotional content?’

“Then every company would spam Reddit with their products.”

‘And then?’

“Then it would stop being useful.”

‘So the thing that makes Reddit valuable for SEO is the exact thing that prevents you from doing SEO on Reddit.’

“Correct. So we need to be strategic about it.”

so every week i round up AI news for my team because most people have better things to do than doom scroll tech twitter. 

making it public now. 

here’s what’s up with AI this week →

Open article: What's up with AI? - 22 January 2026 by slite.slite.page

I'm willing to bet on this. Marketing shifts increasingly offline in 2026.
Text is already slop. LinkedIn is insufferable - every other post reads like ChatGPT with a hustle grind veneer. The comments are worse. X had an epidemic of AI-generated articles flooding the timeline in December. We've learned to read online text the way we listen to a politician's speech: with a grain of salt and the assumption that someone, somewhere, is trying to manipulate us.

But text was just the beginning. Look at Dall-E's first version versus Nano Banana Pro. That's maybe 18 months of progress. AI-generated video is catching up fast too. We're reaching a stage where you genuinely cannot tell if the video you're watching was shot or generated. Give it a few more months.
When that happens, the default assumption for anything online becomes: this might be fake. Not "this is probably real unless proven otherwise" but "this is probably generated unless proven otherwise."

So what's left?

The things that can't be faked. Offline.

I think we're going to see a lot more of what's already starting to emerge: community meetups like Cursor Cafe, Lovable running offline hackathons, brands investing in real-world experiences that require you to actually be there. Even we, at Slite, are starting to go offline - small dinners with Chiefs of Staff to understand their problems, no content play, just conversation.

As signaling on social detaches from effort and our digital trust erodes, offline will only increase.

Ivan Cernja does Growth at an AI startup. A few weeks ago, he built his new portfolio and it made it into my top 5 favourite portfolios from last year. He didn't write a single line of code, didn't wrestle with frames/sections/masking/padding on Figma.

"Claude built ivan.codes in under an hour. But Claude didn't pick Teenage Engineering as the reference, didn't decide a receipt printer contact form was funny, didn't know that friction could be a feature. I keep thinking about what's left when AI handles execution. It's taste. It's having opinions about things that don't seem to matter."

This is the work that gets cited when the 5,000 cultural essays of last year said "Taste is the new intelligence." For once, I don't want to play the contrarian - I agree with all of them. Taste is the new black. I look at my portfolio - a document with ugly-blue hyperlinks - and sigh.

But a few scrolls after Ivan's portfolio, I see an entire marketing team built in Claude Code. 10 AI agents. More skills than my actual team. I close the tab, doomscroll LinkedIn instead - only to be greeted by the beige of slop:

"Build a content factory of UGC creators with Veo 3.1 + Claude Code + Nano Banana" 
"Build 150 SEO articles in one go using Airops right now"

This is my everyday lurking of 2025. I see storytellers being hired for $300K+, I read the 4,000-word Substack essay on taste, I remember Rick Rubin's book, and I feel almost ready to take on the world with my ideas using AI.

Then I go to work, flip the switch, and remember that scaling volume is almost always better on a performance review than scaling quality - which is immeasurable, untrackable, unprofitable meandering.

We all feel it, as marketers. Expected to create art to snatch attention and plug our CTA. Worried about scaling volume - articles, videos, emails - because we know what the upper brass wants. They want something that didn't work before but 100x of it. "If we publish 1,000 AI-generated reels together, we can A/B test and move so much faster."

We've turned ourselves into slop-churning factories. Slop became word of the year in 2025. But there's no term for the opposite of slop. You've never read it on LinkedIn, Substack, or Twitter. Not because good work done with AI doesn't exist. It's because it's outnumbered 100-to-one.

And I don't know if the ratio is changing anytime soon.

taplio

The Battle

Christophe Pasquier

Ishaan G.

Followers

Posts

Engagements

Likes

Performance ScoreChristophe Pasquier's LinkedIn Performance

Performance ScoreIshaan G.'s LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Christophe Pasquier

Ishaan G.

Followers

Posts

Engagements

Likes

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Christophe Pasquier's LinkedIn Performance

Performance ScoreIshaan G.'s LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreChristophe Pasquier's LinkedIn Performance