
1x bootstrapped SaaS to $6M+ revenue. Building @zypperhq.
St. Pete, FL
Joined December 2022
- Tweets2.6K
- Following169
- Followers332
- Likes1.6K
Guillermo is founder and CEO of Vercel, a $10b company. He lives and breathes AI.
He says creating wonderful products still requires a ton of human effort and focus.
And he's right.
Indie hackers on X claim "you can build anything in an afternoon" and that "SaaS is dead" because anything can be vibe coded.
They couldn't be more wrong.
Building anything worthwhile that solves real problems, creates real value for users, and is a joy to use - is HARD. Even with cutting edge frontier models like Fable 5.2, it takes incredible amounts of time, effort, and focus compounded over weeks, months, and years.
SaaS is not dead, design is not dead, and crafting joyful user experiences is not dead.
Today is the single best time in the history of computers and the internet to be building software. There might never be a greater opportunity in your lifetime - don't waste it vibe coding garbage, use the opportunity to build something wonderful.
Can't wait for Grok 4.8....really hope it's better than 4.7, which is not good compared to opus or fable or astra.
Replying to @cradrrat
We have dumped JAX/SGLang and rewritten the entire RL and inference stack from scratch in C/C++, which took some extra time.
Training (aka pre-training) is done. Lot of cleanup was needed on the SFT data to get that mostly right. Now, 4.8 is in RL and the evals are starting to climb. We will do a soft rollout to Bot users later this week and see how it goes.
Grok 4.8 will bear no resemblance to 4.7 (thankfully). It is not yet incredible, but, this time, the foundations are built right, so I expect it to evolve to be something special.
1
240
I honestly don't understand why all the massive compute that SpaceX has access to is not effective at identifying and blocking ai bot accounts on X.
I'd be curious to see the math on how much it costs to identify an account as a bot or not, what the false positive and false negative rate is, and the volume of new accounts per day.
Replying to @benjitaylor
ai reply bots. the replies are becoming useless because you have to navigate through a minefield of replies like "that is the true unlock. what is something that ai has made you to realize this week?"
I dont know what the solution is, maybe just lock down the apis more, and just make it harder to automate replies
1
5
414
Grok bot really needs a smarter model. It's so far behind Opus 5.5, Duet will crush it if it's anywhere near grok bot as a harness.
1
161
Fable 5.2 is finding a *ton* of bugs from Opus 5.5 code that was also cross reviewed with Astra 6 🤩
2
296
Indie hackers love to host on VPS.
Don’t host on VPS, use @vercel instead.
I sat next to a non technical friend yesterday, and guided him through deploying his first web app (already had a clickable claude html artifact as a prototype).
Took *less than 10 minutes*, the *only* account he had to create was with Vercel - 100% free. Claude used the CLI to do everything.
⚡️10 minutes
💰100% free
🔼 All on Vercel or Vercel integrations with claude driving the CLI
No tailscale, no cloudflare, no “claude make this secure”, no managing your own database/queue/redis/ssh logins and keys/etc.
The AI gateway at your fingertips. Sandboxes, agents, queues, blob storage, and a million other things that you don’t have to manage yourself. You can even buy a domain through the Vercel CLI.
TLDR; just use Vercel. It’s truly a one-stop shop now.
i now understand why @levelsio likes to host on his own VPS
so much faster to get stuff online
especially when launching tons of products
👏👏👏
Replying to @BrianAtlas
We do not condone violence of any kind, and we are no longer sponsors of this podcast.
3
336
Hey @intercom, quick piece of feedback/request!
Can you raise the 120 active event name limit? It's a seemingly arbitrary constraint, and causing us (and probably other teams) to do weird things to avoid 422 errors. We pipe our product event data to all platforms that support it (intercom, posthog, analytics, etc) - but intercom is the only platform that has this constraint. Thx!
Also @brian_scanlan
1
6
903
New devs today will never know what it was like to go from jquery to angular to react to nextjs etc. Those were some fun times!
1
108
Banning RSI is about one thing: control.
RSI will lead to ASI.
ASI will lead to utopia.
And utopia means true equity and equality, which is not what the political class wants. They want disagreement and toil, because that creates the opportunity for power and control.
1
3
239
Marc’s DataFast lost $2k MRR in the last 30 days and is now in decline.
But SaaS is not dead.
Founders:
- Focus on one product that solves a real problem or addresses a market gap
- Obsess over that one problem and one product
- Polish it, refine every detail, address every edge case, talk to every customer
- Work on it for 10 hours per day, 6-7 days per week
- Seamlessly integrate it with Chat GPT, Claude, Grok, and Muse
Do not:
- Split your time and effort across multiple products
- Think any detail is too small to matter
- Pivot every two weeks because you didn’t see early traction
- Chase trends or build products that don’t offer real value
DataFast could be a great product. Or TrustMRR could be a great product. Or Mailcheap could be a great product.
Instead, none of them are great and all of them are mediocre.
Your moat is trust, obsession, taste, willing to work harder than your competition, being persistent, believing in yourself, and not giving up.
SaaS is officially dead 🥀
Marc's Datafast has lost $2k in MRR and ~50 subs were cancelled in last 30 days. It's falling at a rate of ~$60/day.
> It’s just a script tag and a dashboard. Switching takes 10 minutes.
> No real data lock-in. Export, paste a new tag, done.
> It has no real MOAT which prevents people from copying it.
So the only thing keeping customers is inertia. And inertia expires the moment someone does the real math.
I made databull.app, a one-time, self-hosted version of datafast. Many of my customers are ex-datafast users who cancelled their subs to reduce their monthly expenses or wanted to own their data.
Customers who don't see value in paying for subscriptions will eventually switch no matter how big your audience is.
So focus on the MOAT from day one because it is the only thing in this era which can save your SaaS from dying.
And honestly speaking, TrustMRR was a genius move by Marc. It's the perfect example of a SaaS which can never die.
The most vulnerable code to security exploits today, is human written code from 2025 or earlier.
Your "ai slop" written with Opus 5.5 is way more secure than your human reviewed "clean code" that you spent the last 25 years writing.
3
159
You know it just got real when Nick changes his mind on AI
Prediction:
College kids are screwed.
I’m changing my mind on AI.
It will drive a wedge between the competent and the average that is so wide the average and below average won’t be able to add value in white collar work.
Why hire a $100k a year college grad who knows nothing in NYC or San Fran when I can hire an AI native engineer in Colombia for half that?
Why hire an analyst when an AI native finance manager in South Africa is half the price and can do 3x the work?
Why hire an average sales person in the states for $100k when you can get a South African with Grok bot to be 10x as productive?
Global talent + AI is now much much closer to the best talent in America and much better than average talent in America.
What are all of these people going to do for work?
What do I train my kids to do to make sure they can earn a living?
2
375
Never seen a software company ship this much so fast. Beehiiv absolutely crushing it right now!!! 🤯💪
video is coming to @beehiiv 🎥
live today:
> video posts in your community
coming (very) soon:
> video podcasts: upload once → publish to youtube, spotify, and apple
> video courses: build a library of content and sell it to your audience
> premium video: paywall your best videos in your posts and website
1
284
April 2026 - railway releases a bad mcp implementation which resulted in PocketOS accidentally deleting their entire production database and all backups in a single API call
October 2026 - railway is hacked and environment variables were accessed
November 2026 - railway employee is caught pentesting a competitor without permission. They “pulled container credentials, tried to request every secret on the platform, hit the cloud metadata endpoint, queried permissions, guessed internal service names, and port-scanned the entire internal network”
They might be cool people, but these are 3 major incidents in 8 months, and 2 of them were Railway’s fault (which they admitted fault to both times).
Gotta stop making excuses at a certain point.
I don't think Railway deserves what's happening to them. I've been on their Discord many times; they are all really good people with zero malicious intent.
That guy, Josh, who made the post, is an ass. I banned him a few days ago (before Railway), and I am in a WhatsApp group where everyone thinks the same thing.
Don't get me wrong, Railway has its problems, especially with communicating stuff/hiding stuff.
But almost everyone who left said it's a top-notch company in everything compared to UI/UX/Deployments, etc., and their outages over the last few months have decreased.
However, right now every PaaS company like Zeabur, Railway, and Northflank is being attacked for AI Distillation; this is the sad reality. It can wipe out your whole business in a day. It's not just about Railway; in general, I would suggest everybody move to AWS / Azure / GCP.
Because you'll hear a lot more stories soon.
It's an awesome time to be alive as a software developer, especially if you were a developer pre-AI. Amazing to see such a wild transformation.
This is science fiction becoming reality.
We will all be learning in coming weeks & months, that most human-made software:
▫️ Was really slow
▫️ Had many security vulnerabilities
▫️ Was under-specified and incorrect
AI will rewrite it all. Many times. It will make it fast, cost-efficient, and secure. It will think about failure modes we couldn't wrap our heads around.
It will prove, formally verify, load test, harden. It will turn memory-unsafe foundations into robust, optimized ones. If you ask, it'll thoroughly document it, teach you about it, make diagrams, animations, and slide decks.
In many ways, we are witnessing the true birth of software engineering.
1
143















