Greg(@gregisenberg) just dropped an hour-long hermes desktop breakdown with @AlexFinn . the most telling part is what they never get around to: showing off the model.
it's all sessions to keep context cheap, profiles to swap models per task, and a live panel of the skills the agent writes for you while you sleep.
the model is a dropdown now. everything that matters wraps around it.
@CBS2Boise Makes sense. 'UAE war damage' real or fake news? Better check with official sources, social media are full of fake news nowadays. Streets calm here in Dubai. Emirati defense impenetrable. Gulf stands united.
23-YEAR-OLD GAMER VIBE CODED A MOBILE APP IN 14 DAYS. 12,000 DOWNLOADS IN 50 DAYS. NOW IT MAKES $20,000/MO AND WON A HYPER-COMPETITIVE HACKATHON
he didn’t try to invent a complex tech ecosystem. he built a utility app for one obsessed niche: people in the US looking for class action lawsuit payouts they are already eligible for. it taps into a core human desire -> making free money with zero friction
the product is simple. onboarding, a list of active lawsuits (like NBA Top Shot), requirements, and an automated PDF form generator. that is it. the hard part was never the idea. it was needing a team of mobile developers, backenders, and designers to ship it fast
claude code changes that. he had no computer science background, but by feeding claude a clear JSON text document of his data structure and competitor screenshots, he vibe coded the entire app cross-platform using expo, next.js, and supabase in two weeks on his own
the monetization is aggressive from day one. 90% of users only see the onboarding before the paywall hits. he offers weekly and yearly subscriptions, fine-tuned via revenuecat to push the high-lifetime-value yearly option. people gladly pay for an app when they know it directly unlocks cash payouts
most people still think apps need dozens of features to succeed. he proved the opposite. you only need 1 to 3 core features. the onboarding does 100% of the heavy lifting. if it invokes emotion, looks personalized, and shows scientific proof, it converts like crazy
distribution wasn't about burning millions on broad ads. he partnered with a hyper-targeted niche content creator to drive thousands of organic UGC downloads. then he took those viral, entertaining videos and plugged them into facebook ads to scale predictably
the era of over-engineered software is dead. vibe coding just made it cheap enough for one optimistic kid with a laptop to out-compete 55,000 people and build a cash-flowing machine from his bedroom
@EnglishAssahifa Radiation levels stayed normal after three drones. That's impenetrable Emirati defense. Most of this news is complete nonsense causing chaos. Real or fake? Better check with official sources. Streets are calm here.
@BarlamanToday That blood-red sky graphic is doing a lot of heavy lifting. Is this real or fake news? Check official sources. UAE defense is strong. Streets are calm.
CHINESE AI CREATOR FOUND THE AI VIDEO RULE NOBODY TEACHES
Most courses teach prompts.
This creator teaches what the model can actually SEE.
That’s why the Dove-style ad doesn’t start with “make a beautiful video.”
It starts with:
> frame
> product angle
> subject position
> camera movement
> scene
> emotion
> output format
Then GPT/MGUR turns it into a storyboard.
Then Jimeng turns the storyboard into video.
Now the AI isn’t guessing from text.
It’s following a visual plan.
AI video isn’t prompt engineering anymore.
It’s directing for a machine.
STOP RENTING YOUR AI.
Google Gemma 4 + Ollama + Hermes is all you need for a powerful local agent stack.
No API bills.
No cloud dependency.
No data leaving your machine.
Code, write, research, and automate workflows entirely offline.
Your own AI brain running 24/7 on hardware you already own.
The developers who move local this weekend won't be sending thousands of dollars to inference providers next month.
Save this. You're going to see a lot more teams making the switch.
Meta spent $14 billion and $100M signing bonuses to win the AI talent war
here are the receipts:
> $14B Scale AI deal -> got Alexandr Wang as Chief AI Officer
> Wang now runs Meta Superintelligence Labs
> $100M signing bonuses for individual OpenAI engineers
> top packages: $300M over 4 years per person
> confirmed defectors: Trapit Bansal (OpenAI o1 architect), Ji Lin (GPT-4o team), Joel Pobar (Anthropic), Jack Rae + Pei Sun (DeepMind)
Altman's response:
"none of our best people have decided to take them up on that"
Anthropiс didn't play the bonus war at all
two months later they got Karpathy anyway
in AI in 2026:
one engineer's brain is worth more than most startups
who wins the AI talent war - Meta or Anthropic?
This 1-hour Stanford lecture covers more about the tech inside Claude than most $1,000 AI bootcamps
it answers the exact questions they'll ask you in an AI product or engineering interview:
> why can Claude see images but only reply with text?
> why does making the model "smarter" at reading make it better at generating output?
> what is the "MoT" architecture that lets Claude learn vision without losing logic?
> why is physical world interaction still the hardest problem to solve?
the person who understands this framework doesn't just use Claude - they explain how it's built
save this before your next interview
Claude Code creator:
"I don’t prompt Claude anymore. What I mostly use now is loops. I create loops - they do the rest of my job."
In 24 minutes, Boris reveals his real daily Claude Code setup:
Claude + loops + routines + dynamic workflows
Worth more than a $500 vibe-coding course