I agree but that simply isn’t a moat anymore. Every 2 weeks, again EVERY 2 WEEKS, a new industry gets repriced.
In the last 3 weeks Anthropic alone decimated SaaS, financial data and cybersecurity. It’s only a matter of months (at most) before some AI releases a no code type wrapper and we have the inevitable NoCodegeddon.
Claude 4.6 is great at one shotting frontend mini apps but hallucinates worse than GPT 3.5 sometimes. Crazy statement.
Thanks to OpenClaw, not Anthropic, creating a scare.
The recent big SaaS stock price crashes were not about “Woah AI is awesome and can build SaaS apps!” Shareholders are starting to doubt the user seat licensing pricing model for these products.
This is from realization that consumers/businesses are buying into AI agents doing core functions of these SaaS products (eg. analyzing data, creating reports), which btw are also deployed by other SaaS.
Exactly. AI is not capable of replacing SaaS…it’s just capable of allowing companies to pay SaaS operators for one seat instead of one per employee. SaaS is repricing for per usage, rather than per seat.
My condolences to his business
Many companies had fired a lot of developers and have been having to hire them back once finding out AI is not replacing the human developers as expected.
For those who were building apps with code before AI ever came along, things are really to the point of hilarious now when you hear some people talk about AI.
Yes, AI is good. Yes, developers use it a lot now. Yes, I like it and use it…
but with AI, it doesn’t mean companies laid off junior developers and are now using just anyone off the street to sit and use AI all day. It’s not how it works. They still use experienced and trained people who know how things work in their field to use AI…it greatly supplements what they already know.
If you had 5 construction workers building a house, and someone invented power tools, you could probably lay off a couple of workers…but the ones you kept would still need to know what they were doing, even though power tools made the job easier.
Anyway, I love AI, just find it interesting how some suddenly believe they’re overnight astronauts or something all of a sudden 
Let’s get the timeline straight.
This wasn’t OpenClaw (and I love OpenClaw). It trended at the end of January. The market didn’t move.
January 30: Anthropic quietly launches a plugin marketplace for Claude Cowork. Not a splashy product event. Just a technical doc update that gets picked up a few days later.
February 3: roughly $300B wiped across SaaS and related tech as investors start pricing in agent-level workflow automation risk.
Then again last week:
February 20: Anthropic publishes a blog post. Cybersecurity stocks drop hard. Within ONE hour $10B evaporates. By the close, closer to $20B+ across the sector.
That wasn’t OpenClaw virality. That happened weeks earlier with no comparable move.
And OpenClaw isn’t a enterprise threat to cybersecurity vendors (although if it posed a threat to them that would be it super ironic).
The fire sale happened the moment Anthropic signaled deeper capability encroachment into SaaS/security workflows.
You can argue the reaction was overblown.
But the catalyst was Anthropic signaling additional capability. Not a wrapper project going viral.
What Anthropic blog post? It isn’t even market leader.
OpenClaw is a now known high level security risk due to how much access it has to a system.
and it just keeps happening! I replied 38 mins ago, so basically ages ago in today’s world or enough time for Anthropic to give IBM a $31B hit and its biggest loss in 25+ years, without even meaning to!
You’re conflating two completely different things.
A tool having major security flaws does not make it a threat to cybersecurity companies. If anything, it reinforces the need for them.
OpenClaw being risky doesn’t logically translate into CrowdStrike and Okta losing tens of billions in market cap. That’s backwards.
Also, the fact that Anthropic isn’t the market leader doesn’t really help your argument.
Yep i found what you were talking about too. Missed this, my bad.
Like you said, it’s a market overreaction. A reaction to Dario’s latest rambling and over hyping.
I never talked about OpenClaw security issues affecting market…
Being killed so many times at this frequency is not good for health 
That’s the reverse of reality. Bubble was already 2 years and millions of dollars in before others started working on it due to its inability to go beyond a skeletal UI scaffolding
I think you’re conflating 2 distinct things. 1) internal write scripts (no IDs) and 2) updating JSON which needs IDs
I haven’t the foggiest idea what it is but technical debt doesn’t explain it (or at least not all of it) because something else is up. What it is I can’t fathom but there’s something going on.
Case in point: after TWO YEARS, Bubble’s native agent can’t even read the issues in the Issues Drawer. It’s native to the app, and it can’t even read the issues in the Issues Drawer.
Note: It asks for a screenshot of the issues drawer but overlooks the fact that it can’t accept images (How does a partnership with Claude lead to the only version of 4.6 that can’t digest images? 
)
Yet BUBS, who is turning TWO WEEKS tomorrow, (since building it began, not since shipping), has been able to extract the issues in Issues Drawer and has fixed 932 of them across a dozen apps over the last week.
BUBS extracts the issues → saves them to the board → diagnoses the actual problem → figures out the solutions (asks for human input when the issue crosses the threshold for the specific project) → executes the fix → verifies the fix took —> add notes of explanation & issue ID to the element / WF.
Since my account is obviously going to be suspended after this post, I will post the BUBS’ live activity ticker where people actually are, like on LinkedIn 
Maybe, I don’t know, but when I built stuff 18 months ago to create new colors, data and styles, it sort of seemed like the same thing, where I create my own IDs and update the JSON and then, like magic, they are in the apps JSON…its almost has if a POST/PATCH are the same type of api calls in Bubble
AI Code and Vibe Coding are a waste of time if not handled by someone with indepth expertise of how to hande the coherent picture / architecture of what you are building. The first couple of itterations may seem solid but when you make changes along the way the A.I. may just as easily be convinced that it’s initial work contains logical errors the way it was convinced it was doing the right thing the first time. Guardrails and checks are everything, and with the vast amounts of code generated, performing quality reviews becomes a risk of routine acceptance for speed and convenience with increasing risk of undesired code injection or vulnerabilities.
What works perfect on the other hand is setting op a Virtual Machine (or something ring fenced from confidential data such as passwords etc.) and the using Claude Code to operate a platform like Bubble.
In an overnight run Claude Captured a significant amount of bugs and inconsistencies on our app (e.g. 12 different types of placeholder for empty values (such as n/a N/a - na not applicable etc.) which you would typically solve beforehand if you would work on the basis of frameworks/ best practices (which unfortunately is typically your lacking expertise when you start using Bubble and runs in some technical debt), but now can be resolved at a low expense with significant accuracy.
Next we are going to ask Claude to create a Bubble app based on the App Framework generated by an extensive back and forth between Sonnet and Opus creating app framework documentation.
No it may not go as fast as directly writing the lines with AI but:
- I have full visual control of the outcomes via the changelog
- I fully understand what was built without reading a line of code
- I’m not at risk of things being created beyond the framework applied by Bubble - so fully relying on Bubble’s tech - no risk.
So for the coming period - I’m more optimistic about the collaboration between AI and technically sound and ring-fenced architecture, fully controlled and managed by experts (i.e. Bubble)!!