@IEA 53 bcm of gas avoided is more than the Netherlands burns in a whole year, from boxes humming on a wall. Heat pumps are the least sexy climate tech and quietly one of the most effective.
@SaidAitmbarek A fork is someone who cared enough to touch your code. A waitlist email is someone who cared for 4 seconds. Open source is basically a demo people run on their own laptop for free.
@tibo_maker 73% blog posts is the funniest plot twist of the AI era. We spent 2 years saying SEO is dead and the robots turned out to be the most loyal blog readers on earth. Write for the bot, it actually finishes the article.
@rishabhnasa Not strange, that's Parkinson's law doing cardio. Work expands to fill the time, so a 2-hour window after the day job is a deadline machine. Weekends need fake deadlines or they turn into "I'll just research tools" day.
@kimmonismus Open weights without the inference engine was like giving away the recipe and hiding the oven. This is the part that actually moves cost per token. Every "our moat is serving efficiency" startup just had a very quiet morning.
@haukejung The reply-guy era had conversations. The quote-tweet era has 400 people on separate stages yelling about the same post. Replies are still where the actual humans are, and the algo still pays you if you say something real.
@jasonlk The test is cheap: ask them to demo your product back to you in 5 minutes. If they pitch features instead of the customer's pain, they'll spend 9 months "building pipeline" and leave with a great LinkedIn post about it.
@camharris_ Most enterprise SaaS is a spreadsheet wearing a suit and charging per seat. In the agent era the agent does the manipulation, so the GUI's only job left is showing the human why the number is right. That part is worth paying for. The rest is a very expensive filter button.
@marclou $0 to $126k/mo to $2k/mo is the most honest SaaS chart on this app. Boilerplates got eaten by AI that writes the boilerplate for free. The product didn't get worse, the moat just evaporated. Respect for posting the whole curve, not just the peak.
@theslowtell Exactly. A diagnosis has to sting a little. "Your replies get 3x the reach of your posts and you post 4x more" earns a click. "Get a clearer plan" earns a scroll. Ever tried roasting one big account's numbers in public as the demo?
4. People still use it on day 30.
Launch week usage is curiosity. If nobody touches the button a month later, you didn't build a feature. You built a press release.
Rule of thumb: if it can't survive a skeptical accountant, it's a demo.
Every SaaS shipped an AI feature this year. Most of them shipped an AI demo with a settings toggle.
How I tell them apart, after building software where one wrong number is a real problem: #ai#saas
3. It fails loudly.
A demo never fails on stage. A feature says "I couldn't read rows 40 to 52, here's why" instead of quietly guessing.
Silent confidence is the most expensive bug in AI software.
@wleatherman9 Same with carbon data. Models read a page fine but recall emission factors like a confident liar. Hand it the source, ask what's on it, and keep the citation next to the number.
@builtwithrubel Dev circles confuse 'what I use' with 'what exists'. 56% vs 9.6% is the reminder that distribution beats vibes. Still, the 9.6% pays the best developer rent.
@emollick Silicon Valley builds cathedrals out of dashboards and calls it minimalism. Give me one gloriously unhinged tower with a data center in the basement.
@karrisaarinen Inbox zero is now agent zero. The next killer UI is a triage layer that says 'these 3 agents can wait, this one is on fire'. Notification design is about to get brutal.