Naive take: rather than chasing smarter models, why aren't AI labs all-in on infinite context (5M+ tokens) despite 'lost in the middle' & compute hurdles?
Imagine onboarding LLMs with your full digital life—emails, SMS, online activity, pictures, google drive documents, social media activity, contacts - literally every bit of informatin available on you - all in-context.
I feel this could 100x anyones productivity w/ existing models, mimicking brain-like reasoning where life's context drives decisions. In my exp, models underdeliver mostly from context gaps. Thoughts?
day 19 of chasing 10k MRR: just landed my first paying customer.
the journey has been anything but smooth—lots of failure, one small win.
here’s my full breakdown (with numbers): what worked, what didn’t, and how I got here 🧵
@ErnestoSOFTWARE recently learned this lesson the hard way.
also learned you should start marketing at the same time, or even before, you start building the app.
10/ tiny wins are everything when you’re building from zero. Learned more in 2 weeks than in months of planning.
if you want real-time lessons from the trenches (failures and all), follow along. more experiments, more learnings coming soon.
this sounds eerily similar to Google's original 'organize the world's information' mission. started with pure algorithmic ranking, but economic pressures inevitably crept in. Hard to see how AI platforms won't eventually introduce 'premium placement' or 'featured content' tiers once publishers start demanding better ROI.
@fortelabs agree. also morally repugnant to deny whole regions access to AI.
was recently in Venezuela where AI is only accessible through VPNs or paid plans.
countries without free AI access are being economically harmed and put at a massive competitive disadvantage.
@alexgraveley I see hundreds of breakout apps - just not in massive markets. AI coding is letting specialists build for their specific niches, and these focused apps are way easier/cheaper to scale than another generic AI platform.