Own your mind. Own your work. Own your stack. Ex-agency owner, 20+ yrs in tech. Building apps, local AI and a side of travel. Exit the default.
Get My Help π
Joined September 2021
- Tweets414
- Following158
- Followers100
- Likes424
Pinned Tweet
This is what going solo at 40 feels like.
Iβm Aaron. 20+ years in tech. Ran a marketing agency for 8 years. Built apps and websites as a freelancer along with almost all of our agency systems.
In August, my business partner left our SaaS startup.
Now itβs just me, rebuilding on X while traveling across South and Central America.
What Iβll share:
β build lean with AI
β run AI on your own hardware
β live and work from anywhere
What are you building, or trying to get built?
Lets connect and grow together!
29
20
1,063
Two Haung's don't make a right
JUST IN: Nvidia CEO Jensen Huang had another affair with an $NVDA executive, NYT reports.
1
2
30
A secret Grok link is not a lock.
This report says 35,000+ apps were left open because the only protection was a web address nobody was supposed to guess.
If your app has a page "nobody will find", put a login on it today.
βΌοΈBREAKING: Grok Build has left 35,000+ of its users' apps open to anyone, including apps from SpaceXSI.
The open apps held 3,784 medical exams of workers and a Tesla technician scorecard, plus a Navy contractor's roster and SpaceXSI's own sales plans.
The "secret" words in each app's link sit in its web address, which DNS logs record, security firm Claroc found.
Cursor, which SpaceXSI bought, says apps are private by default. Claroc's tests show otherwise.
SpaceXSI pulled its 24 apps after Ars asked. Most other data was still exposed when Ars published.
1
30
You donβt need the cloud for confidential work.
I run DeepSeek v4.1 Flash on 2 DGX Sparks. 30-50 tok/sec, multiple streams at once, free inference.
Anything confidential or not ready to be public stays on my own hardware.
Whatβs one thing youβd never paste into a cloud AI?
1
98
Tasked Fable 5.2 @claudeai with creating a promo video to promote itself
This was "two shotted"
What do you think 1-10?
1
60
Most people only dream about sending this email.
I sent it after 14 years in corporate.
Iβd already given my two weeks. A few days later, I walked out the door past managers and friends.
Steady pay. From the outside, the perfect setup.
But around 4pm one Monday, it hit me. I wasnβt tired. I was watching my own ideas die, one meeting at a time.
So I jumped. And joked I might end up foraging for food.
I was only half joking. Iβd be living off savings. And the money scared me less than what people would think.
But a paycheck can disappear. What you learn doesnβt.
Thatβs what I took with me. It became 8 years running an agency, 20+ apps built, and now a solo business I run mostly from Latin America.
And the β$100/hr ;) jkβ at the bottom?
Wasnβt a joke for long.
What would you walk away from if nobody was watching?
2
79
If this chart of Qwen4 27b is accurate, it's game over
Imagine having Opus 5.5 on your desk...
6
618
Is this for real?
Qwen 4 is coming Now & Beats GPT-6, Fable 5.
- 27B variants land first.
- leak say the first samples are already approaching Astra/Opus 5.5-level quality.
- and Alibaba confirmed Qwen 4 will heavily use RSI (Recursive Self-Improvement) week ago.
- across the entire lineup Max, Plus, Flash, and 27B.
- Target window: today or tomorrow may-be
- Open-source in November gets very interesting, after long time silent.
- The gap is closing,
1
4
64
βArenβt you scared to go to South or Central America?β
The answer is a resounding no:
- Itβs no more scary than much of the US
- Thereβs no lack of McDonalds π
Of course it depends on what part you visit, but Panama is super safe and a fairly wealthy country with the canal.
Cost of living for most things is still less than the US, especially Uber rides and services so your US dollar goes further.
While itβs not as cheap as Colombia/Ecuador - itβs a great entry for any expat trying to adjust to life outside of the US with still maintaining a close proximity.
Whereβs your favorite place to travel or live?
3
85
I have to agree with Andrejβ¦
I trained my 2B model for just for 3 days and many of its specialized security code review tasks are already running better than Deepseek 4.1 Flash 1017
Smaller models makes much more sense from a standard hardware perspective bringing processing to edge devices.
I am lucky to have 2 DGX sparks with 256gb unified memory and a MacBook Pro with 48gb but the majority is working below 32gb of memory or vram.
A lot of the frontier models are a waste and overkill for many specialized tasks that businesses need.
The future is local AI
Andrej Karpathy:
βA 1B model can be much more effective than most current LLMs.
90% of current models waste a ton of capacity and money on remembering stuff that doesnβt matter.β
here is his 5 core thesis about Small LLMβs:
1. the intelligence was never the knowledge. Strip the memorized internet, keep the reasoning. Whatβs left is the cognitive core, and itβs small.
2. the core might fit in ~1B params. Maybe less. βPlenty of room at the bottom.β Weβve been paying for a library when we needed a brain.
3. size stopped being the bottleneck. What makes an agent work is the tools it can call and how tight the feedback loop is, not parameter count.
4. intelligence lives in the system, not the weights. A raw LLM is a CPU with no OS. Small core + good harness > giant model alone.
5. you can own it. A model on your hardware is one no one can rate-limit, price-hike, or deprecate on you.
in 12 months the gap between teams shipping on-device and teams paying per-token frontier bills wonβt be capability.
TwIL an 8.8B model at 90%+ of frontier isnβt a benchmark flex.
Itβs the thesis Karpathyβs been repeating all year, made real.
study more on this below β
1
2
85
If you want to follow my AI model training go here: nighttrain.ai
2
42
