Riches are in the niches. Long-tail of AI.
The race to the bottom will only be for the #buildinpublic indie hacker stuff that's fun to build
Most successful SaaS products are in obscure niches solving problems indie hackers have never heard of
You can still build stuff that's not analytics or wrappers and be successful
But you do need specialised domain knowledge and market affinity
Which you always needed for SaaS tbf
Not sure what to make of it, but new software products' lifespan had a 36% drop in the last 12 months.
Is it due to vibe coded abandonware, distribution challenges, state of economy, mixed factors or something else?
saashub.com/product-graveyar…
narrator: "in this episode of The Shed.."
There are 900+ software categories listed on Capterra and G2. 99% of them haven't been replaced (some will never be) by bespoke AI, while everyone on X is trying to build/copycat the same few tools. Amazing time to live in when code is not a bottleneck. Think outside the bubble.
We're getting to a thin client era. Future of 15+ inch and dGPU laptops is bleak for dev work, when you no don't need big screens for IDEs and any small device (phone, tablet, small-laptop) is enough to manage cloud AI agents. I use Cursor agents (PWA) from my Android phone.
More I chat with AI, more I find it gives general views instead of deep insights. More I use agentic coding, more I realize it needs frequent direction/oversight. How do others proclaim one-shots or loops -- big budgets for top models or blindly accept output?
Not that I use Windows anymore, but while you're at it @Microsoft please stop placing the useless Copilot key instead of the good old Ctrl key on OEM laptops. Lot of us use it for fast moving across words in everyday work.
Whomp whomp.
The Copilot PC brand is officially dead☠️
Microsoft and OEMs have backed away from the label after a messy rollout and months of confusion. Here’s what happened, and what’s next for Windows 11’s AI strategy.
Also, we bet none of you are actually surprised.
Full story in a new report from @zacbowden👇
windowscentral.com/microsoft…
me: hey AI, that didn't/won't work, how about this..
AI:
This is the classic Combinatorial Explosion problem..
You've landed on Delegated Relevance pattern..
You have hit the exact Precision vs. Recall Trap..
What you have described is a Hybrid Deterministic Architecture..
Red pill for AI psychosis..
The more AI news you consume, the more you start feeling like the world is moving 100x faster than you. Every day there is a new model, new benchmark, new agent, new this changes everything thread. Slowly you stop learning because you are curious and start learning because you are scared of becoming irrelevant.
It happened to me too. For the last 2 months, I’ve barely followed any AI or tech updates. No daily model wars, no chasing every release, no trying 20 new tools every week. And when I look back, I can hardly find anything that fundamentally changed what I should be doing.
That’s when I realised most people are not behind in AI, they are just overloaded with AI news. Stop treating every update like an emergency. Build more. Think more. Master fundamentals. The truly important changes will survive the hype cycle and reach you anyway…..
Codex Cloud seems slower than Cursor Agents (also using Codex 5.3) and needs polishing, like file references in its response linked to the editable file lines for context, although Codex has a nice 'Start task' button right next to each item.
Claude Code web any better?
Okay I built it!
🍰 Infinite Slop
levels.io/infinite-slop
An infinite and interactive AI generated live stream of slop that goes on forever and ever
Anything that you write in the chat is generated next and AI will try to connect it to the previous video so there's an actual sequence and story line
Idea by @marcantoinefon and @rehan_shei's original stream so I made it
It'd be very expensive to run this so @fal very generously sponsors this
They also fine-tuned the model that makes this possible for the first time, making Minimax H3 50x faster and as "Max" and it can now generate videos faster than you can watch them, very cool!
Let me know what you think!!!! 😊
I think even ageism will be curbed, with AI giving boomer/genX coders (like me) and knowledge workers greater productivity and an extension in the industry just in time.
Getting older may slow the fingers a little, but agentic work may free up the mind.
This point from @pmarca has stuck with me ever since he came on the pod, that AI came just in time to save us from slowing growth and population decline
"Getting Real" and Basecamp were inspiring as an indie founder WFH.
Switching from Windows, I can tell there's ton of gate-keeping and friction around Linux distros. Omarchy is trying to break it, and it needs such force against proprietary crap.
More power to DadOps Linux!
No one I know outside of the X bubble is cancelling software and rebuilding internally. I read few tried, but returned back to managed/cloud services due to excessive time spent and lower quality. AI codegen is still a great tool, not an end-all (yet).
AI code gen isn't making teams work less, but it's making them ship more (can't say about quality or slop). Linear's data shows coding-agent teams went from 21 to 65 PRs/week in two years, while traditional teams only went 8 to 10.
linear.app/data
Anthropic will embed invisible watermarks in Claude text under EU rules. I guess this will become the norm with all big AI models, even beyond the EU, which may be good to deal with slop (copy-paste detection or pre-ingestion filtering for training data).
interestingengineering.com/a…
Sceptical of KPMG's report:
* 81% of APAC companies say AI is delivering value
* 4 in 5 say AI will remain a top investment priority
but
* Half have delayed or scaled back AI rollouts due to costs
* Only 20% have successfully put AI into workflows
kpmg.com/xx/en/our-insights/…