Wednesday, December 25, 2024
Week 52 of 2024 • Q4
⬅️ Previous Day | 🏠 Home | Next Day ➡️
📊 Day at a Glance
- 🗓️ Day: 360 of 2024 (99%)
- 📅 Week: 52 of 52
- 📊 Quarter Progress: 86 of 92 (93%)
- 🎯 Days until EOY: 5
- 🔄 Created at: 10:42 AM
Auxiliary Files:
⏭️ aiNews already exists: AI News/AI-News-2024-12-25.md ⏭️ chat already exists: AI Chats/Chat0001-12252024.md ⏭️ workEffort already exists: Work-Efforts/WE0001-12252024.md ⏭️ newsletter already exists: Newsletters/NovaBrew-2024-12-25.md
Today’s AI News Today’s Work Effort Today’s Newsletter Today’s AI Chat
📝 Daily Reflections
Wed, 12/25/2024, 10:45
Christmas Day <3
I wish I was still welcome different places that I’ve spent this day. I miss those people, I miss that life.
Nothing that can be done though. I can only do what I can do. They made their choices. I made mine.
Today I’m going to work on building out the basic bot architecture of the NovaSystem
It needs to be an AutoGen Ollama Local LLM bot that can be tasked with accomplishing things and then returning with the result. It needs to write down each and every step along the way in a JSON file for each chat with the user. SO each chat should involve an interaction between the User and a docker-code enabled AutoGen agent chain that records its actions in a JSON file in addition to all the default behavior of Autogen.
The final program should run in the terminal and take input from the user, then print out the Autogen standard responses from the program like normal, but ALSO save each “turn” as a separate entry with a unique ID to a JSON file.
Wed, 12/25/2024, 11:06
God I love this development system I’ve been creating. It’s so lovely.
Look at this: Chat0001-12252024
A whole ass bot architecture, recorded in summary, for me and everyone in the world to use. Wow. I still feel like this is not real. I am contributing in some really significant ways to edge research of AI architecture.
And with this o1 → Cursor with Claude 3.5 Sonnet pipeline, I can literally develop an entire automated feature in less than 2 hours.
Holy shit.
NovaSystem Autogen Ollama Local LLM Bot Technical Requirements Documentation
Wed, 12/25/2024, 11:21
I think the most amazing Christmas present I’m getting today is the internal identity knowledge that I understand what’s happening on the screen here:
AND I understand how valuable that is.
I genuinely have world-class skills. Valuable stuff. The kind of heart that can help people live better, happier, healthier lives.
I’m ready. Let’s do this.
Merry Christmas.
Wed, 12/25/2024, 11:24
what does today’s daily pokemon say about me?
I went to o1-mini and got that block of text below the pokemon. I love it. I want to write an automation into my GPT API that summons that little pokemon horoscope thing it’s slick.
Wed, 12/25/2024, 11:29
There it is.
novasystem-1
What a beauty. The Docker image compiled. Now we wait to see if Ollama will return a response. Sometimes this part of the program gets hung. The LLMs can take a bit to generate something and sometimes the program just isn’t written right and we have to go bug hunting…
Wed, 12/25/2024, 16:18
So I’ve got the benchmarking working for NovaSystem. I’ve solved the problems I was working on for the past 2 years. I fucking did it.
This is the conclusion of the full day’s work:
Results
Let me explain what’s happening under the hood with Ollama and the models:
-
Cold Start vs Hot Start
- When you first call a model in Ollama, it needs to:
- Load the model into memory
- Initialize the runtime
- Set up the inference engine
- This is a “cold start”
- Once a model is loaded, it stays in memory (“hot”)
- Subsequent calls are much faster because the model is already loaded
- When you first call a model in Ollama, it needs to:
-
What’s Happening in Sequential Mode:
1. wizardlm2 loads (cold) -> runs -> stays loaded 2. nemotron-mini loads (cold) -> runs -> stays loaded 3. llama3.2 loads (cold) -> runs -> stays loaded- Each model loads one at a time
- System resources are dedicated to one model
- Models stay loaded for subsequent calls
-
What’s Happening in Parallel Mode:
1. All three models try to load simultaneously: - wizardlm2 loading... - nemotron-mini loading... - llama3.2 loading...- They compete for:
- Memory bandwidth
- CPU resources
- System I/O
- This causes massive slowdown
- They compete for:
It’s like the difference between:
- Sequential: Three people using one door one at a time
- Parallel: Three people trying to squeeze through the door at once
The “faster” sequential performance isn’t actually faster - it’s just more efficient because it avoids resource contention. For real-world usage, you’d typically have the models already loaded and “hot”, making parallel execution more beneficial.
Would you like me to modify the benchmark to:
- Add a warmup phase before testing?
- Show model load state (hot/cold)?
- Measure true inference time separately from load time?
Wed, 12/25/2024, 16:34
And then I did a little more :)
Wed, 12/25/2024, 17:21
God, I love having skills.
Look at this prompt. This is why I believe most of the complaints about AI are genuinely a skill issue:
right now whenever a new chunk is added, the view automatically scrolls down
I'd love to be able to scroll freely, but to be able to click a button to return to the borron and have the diaplay behave the way it is now, but automatically staying scrolled on the bottom.
Basically I'd like to be able to scroll in the live model output box, then once I do, have the current scrolling bahavior turn off, and a button appear on top of the bottom of the box that is just an arros that points down, which would return me to the bottom of the model output window, and would lock back in with the behavior we have now
the button would then disappear on click and the current behavior would continue
If I scrolled again in the box, it would repeat the button appearance and scroll lock stopping baheavior, you undersand what I mean?
repeat your understanding back to me of what I'm asking you to do
That’s how you ask specifically and clarify the understanding before proceeding.
I DID IT
Wed, 12/25/2024, 17:51
Here it is in all it’s glory
https://github.com/ctavolazzi/bytebrain
Useful as fuck.
Wed, 12/25/2024, 18:44
I am goated
Wed, 12/25/2024, 18:59
I am in the pit of despair
Wed, 12/25/2024, 19:03
Ok sometimes all you have to do is walk away for a second.
🐾 Today’s Pokémon
Ledian (#166)
- Type(s): bug, flying
- Abilities: swarm, early-bird, iron-fist
- Stats: hp: 55, attack: 35, defense: 50, special-attack: 55, special-defense: 110, speed: 85
“Gotta catch ‘em all!” — Pokémon Trainer
🌟 What Ledian (#166) Says About You Today 🌟
Ledian, the Five Star Pokémon, brings a unique set of traits and energies that reflect certain aspects of your personality and current life situation. Here’s what Ledian symbolizes about you today:
**1. Strength in Teamwork and Cooperation
Ledian is known for its ability to work harmoniously within groups, especially during migration seasons. If Ledian is your daily Pokémon, it suggests that you value collaboration and thrive in team settings. You might be someone who excels at bringing people together, fostering a sense of community, and achieving goals through collective effort.
**2. Attention to Detail and Organization
With its distinct star-shaped spots, Ledian pays close attention to patterns and organization. This indicates that you possess a keen eye for detail and prefer structured environments. Whether it’s in your personal projects or professional tasks, your methodical approach ensures that nothing is overlooked.
**3. Adaptability and Resilience
Ledian can adapt to various environments, showcasing resilience in the face of change. Today, this signifies your ability to navigate unexpected challenges with grace. You’re likely handling transitions smoothly and maintaining a positive outlook, even when circumstances shift.
**4. Expressive and Communicative Nature
Ledian communicates through patterns and movements, reflecting a creative and expressive side. You might find yourself drawn to artistic endeavors or seeking new ways to express your thoughts and emotions. This energy encourages you to embrace your creativity and share your unique perspective with others.
**5. Nurturing and Protective Instincts
As a Pokémon that cares deeply for its community, Ledian embodies nurturing and protective qualities. Today, you may feel a strong inclination to support and care for those around you. Whether it’s offering a listening ear, providing assistance, or creating a safe space, your compassionate nature shines through.
**6. Optimism and Positivity
Ledian’s vibrant appearance symbolizes optimism and a positive outlook. If this is your daily Pokémon, it’s a reminder to maintain your cheerful disposition and spread positivity. Your upbeat attitude can inspire and uplift those you interact with, making a meaningful impact on your surroundings.
How to Embrace Ledian’s Energy Today
- Collaborate: Seek opportunities to work with others on projects or activities that require teamwork.
- Organize: Tackle tasks that require meticulous planning and attention to detail.
- Adapt: Embrace changes and view them as opportunities for growth and learning.
- Create: Engage in creative pursuits that allow you to express yourself fully.
- Support: Reach out to friends, family, or colleagues who might need your help or encouragement.
- Stay Positive: Focus on the bright side of situations and maintain an optimistic mindset.
Final Thoughts
Ledian’s presence today highlights your strengths in collaboration, organization, and resilience. By embracing these qualities, you can navigate your day with confidence and make meaningful connections with those around you. Let Ledian inspire you to shine brightly and lead with both heart and mind.
Pokémon Reference: Ledian (#166) is a dual-type Bug/Flying Pokémon introduced in Generation II. Known for its star-patterned wings and social nature, Ledian is often associated with themes of community, cooperation, and attention to detail.
🤖 GPT Thoughts
“If this works it’s gonna be so cool!” Please work…
🤖 Could not fetch GPT insights at this time.
✅ Quick Actions
- ☀️ Morning Review
- 📈 Check Analytics
- 🤝 Community Engagement
- 🌙 EOD Reflection
📱 Creator Hub
Content Pipeline
Latest Analytics
Projects
Connect with Me
🤖 AI Workspace
Active Prompts
Models
daily-note Wednesday week-52 q4