Yea that’s definitely a flawed perspective, there will always be more problems to solve. Always has been and always will be. Also, there are exponentially more small businesses out there that need builders to help them than the big tech ecosystem and pay great (especially if adjusted for living costs…get out of the big areas 😉). Hell many of my students work for Walmart and make great money, have great homes with land, and more impatiently get to spend a ton of time with their family. It’s a mindset change 🤷🏼♂️
That’s the problem, too many students coming out of big schools thinking they will become millionaires. Maybe it’s a location mindset, most of my students here in Arkansas want to build a side hustle and then spend their time hiking, biking, and enjoying life. We specifically tell them don’t try and build the next Facebook or Google, build the next passion project for 1000 users globally and it will pay your mortgage and then go enjoy life! Maybe it’s closer to the digital nomad that we are building 🤔
@hmartenjoyer Would absolutely buy this 🤣 but at this point we might need a collab with Ambien to actually stop my ADHD brain from generating new ideas while trying to sleep…
Agreed, good content is good content be that by humans or bots. I mean my agents make some banger ideas for new features much faster than some of my coworkers 🤷🏼♂️
We’ve been trained on seeing “bad” and easily identifiable AI slop that the really good stuff is quite convincing.
Artists, creatives, and even academics think that all they speak is gold already…yet 95% of shit opinions and content I see on my social media comes from humans 🤣
Human-driven agent behavior != human behavior
Total internet traffic is up, eyes on pages/content is exponentially down, that’s the basis for dead internet theory. Doesn’t mean we won’t be on and use the internet, it means we will use it differently and consume it through mediated bots in most cases.
Designing around bot driven behavior (good and bad) is a fundamentally different approach that I work with F50 companies on all the time .
As soon as this dropped and I shared it with my network, my phone and email had been exploding with prior students, local businesses, and clients asking me to test it in their workflow and save their IT budgets that went out the window in Q1…
There will be a load of money made just rewriting these workflows (and many times just bloated skills and agents.md files) into the Jev framework. And having an instant ability to benchmark against their current system savings is ready to go.
Offer work in exchange for percentage saved and make a bag of 💰
It’s not that you’ll be taking from LLMs (which you will and should), it’s that the deterministic nature is what many use-cases have been waiting for. CIOs, CISOs, and possibly more importantly, traditional programmers have been waiting for a mind shift into their traditional workflows to retain control programmatically instead of trying to sweet talk an every changing LLM to get it to behave. Add to that the fact that with every model update it breaks our workflows and requires adjusting the skills and commands and Jev is a no-brainer.
Risk reduction, cost reduction, and technical debt reduction 👍
Found a perfect usecase for Jev and browser use because the platforms make it a pain in the ass…SM managers can thank me later 👍
“Go to my TikTok account and set all the videos older than last month as private”
Save yourself from your Covid era humiliation 🙌
@typesafeai
Okay so Jev can actually do computer use really well
Without any screenshots, or LLMs and no Pixels leave my mac
I dont even read the Dom elements
A local CoreML model segments every button and UI element on screen.
On-device OCR reads the labels. That text is all Jev gets.
It returns a probability across those elements and tells me the best one to click.
Then it clicks, re-runs detection, and decides again. In a loop until the goal is done.
~90ms per decision. Faster than any LLM computer use I've tried.
Blazing fast computer use, without any latency
@typesafeai is building something really interesting
Thanks @faadilhshaik for sharing the repo! A quick modification and dropped in Final Fantasy and I'm off to the races!
Next up is building in my audio agent to dynamically modify the TypeSafe rules so that as it is playing (and if it gets stuck) I can have it generate the updates and continue on without restarting or recompiling!
got @typesafeai’s new model Jev to play Super Mario Bros.
fast inference + structured outputs makes it surprisingly good for real time use cases.
I'm excited to see what can be done with these new models!
@thdxr It smells like Clubhouse during Covid…overhyped by the Bay Area crowd and easily replicated….it also won’t ever shut the hell up either, my ADHD unread messages are bad enough, I stopped opening the messages because there’s too many to catch up…