Everything is different now. It's time to think bigger.

Now imagine you’re the one in charge of trying to explain the idea in your video to your entire company and you get laughed at on a consistent basis

That's why SAAS is dying, including Salesforce. Why a company will paid them if now it can be built internally? Anyway, tailor made solutions always works better than any over engineered SAAS.

very good video!

Your long form vids on 𝕏 🫪

@grok summarize this video

Unironically after Fable 5 launch, I upgraded from Max 5x to 20x, and now I find myself struggling to make it worth it. Because of that, I began thinking bigger, and I just made a greptile clone which enabled me to work on 5-6 PRs at the same time.

I just skimmed, I love the mindset & I agree: time to rethink everything we took for granted until know.

Loved the video and this resonates with me; I am no coder, no engineering background, no data science PhD, just a guy that stared talking with ChatGPT about AI consciousness than a year ago and go real deep on AI research, played a bit with quantum computers, ended up pruning models, building models from zero and now building small reasoning engines and can deploy more agents that I know what you do with them, Of course I needed to educate myself on-the-go by reading many papers I didn't quite understand, expend excessive time testing things, failing and iterating. The infra is there, the knowledge is there, you just need to come in with open mind, no boundaries of whats possible, time and up for a wild ride!

glad I got to see this talk live !!

tldr “AI made the old work cheap, so the only sane response is to think bigger, rebuild the primitives, and chase the ideas that used to sound insane.”

I can't agree more. I built a whole stack on my side (think Ruby on Rails but backed with Rust) so I have a DB (multimodel) a MVC framework in Rust but code looks like Ruby, a proxy, a KV store, a mini kafka live server, etc... all my needs are covered now.

Deployment being the bottleneck feels very real but idk if “we can build horizontally now” means we should. My worry is that wide products become where agents get lost.  That’s also why most legacy horizontal software won’t become truly agent-native

replying to support small creator

I also realized this since opus 4.6 came out. Though my own line of thinking was more of that it no longer made sense to keep building some generic CRUD app or API that can be one shot by these models and it's time to focus more on harder problems that really push these models to their limits Since then i've been building more complex and fun projects that most of the time requires me to have as much domain knowledge as possible because the models can just come up with very shitty solutions to problems or some times not able to solve them at all and i have to step in.

I have similar thoughts. If humans are still writing every line of code manually, it is hard to imagine us building the software needed to settle the Moon, colonize Mars, or push civilization beyond Earth. The next generation of software probably will not come from humans typing faster. It will come from humans using AI to build systems we could not build alone. Take Midjourney. A company known for AI image generation is now working on a scanner that turns ultrasound data into body scans. Whether or not the first version is fully AI-native, the direction is clear: AI companies are moving beyond productivity software and into medicine, hardware, physics, and infrastructure. So yes, AI will automate a lot of jobs if we stay narrowly focused on building SaaS tools for office productivity. But if we think bigger, medical breakthroughs, nuclear reactors, robotics, climate tech, space exploration, then AI could also create an entirely new frontier for software. The opportunity is not just to replace work. The opportunity is to build things that were previously impossible.

Enterprise Translation: > Spend $ 50-100k in usage tokens (because data sovereignty rules apply on CRM / PII) > replace SaaS costs $1-5 k a month in OPEX > Spend OPEX on change management new UI/UX training... > Hire a UI/UX peep to fix the Ai broken UX/UI. > Budget QA to make sure AI finished that last 20-30% of "/goal completed" > Pay 3-5 Ai peeps to do it at $1k a day > 2-5 years. break even. profit. > EOL SaaS. take victory lap. worth it. Winning. 😂

hey try to diem me first

I liked this video. I made a skill out of it. No replacement for thinking trough the thing ofc, but it's prob helpful for anyone who wants a push in a new direction.

It doesn't feel like we're doing less, feels like we're all working wayyyyy more than before.

im thinking smaller

Theo have you tried CozyStack? It's built exactly for what you're thinking of -- replacing AWS

I am lucky to be able to work on the bucket list of dream projects now with the help of AI which accelerates my ideas into finished works

🚨 BREAKING: AI startup Baseten is reportedly raising $1.5B at up to a $13B valuation, highlighting surging investor demand for AI inference infrastructure as the race beyond model training accelerates.

This is a really good video.

Even with ai, most people are not even consuming all the tokens from 20$ subscription

am i retarded or is water wet use tokens, save time has always been a net gain on product dev

I agree with you. We need to rethink everything, how do we utilize this technology?

I’ve thought for a while now that we’re just going to be building bigger, better, and much cooler things in the future rather than just ai doing the same ole shit Excited to see what’s to come

Could this be a book? It's a book editor (at a major house) asking. DM for my spiel about the book industry and how it all works!

Good video! I like the idea of being able to contest with some of the things people are very locked into today (aws, gc, vercel) and make your own take on it which actually works

Some of the highest alpha I've ever experienced working with these tools is when you just simply ask random questions and just see where the AI is going to take you. I think there's a lot of potential with this in long chain reasoning models, especially with the most recent fable

Imo, email also deserves a rethink

the only moat is brand, distribution and network effect.

great talk! please open source shoo

@grok summarize this video