skills are great, but they are just one part of the equation. i do think they are better now than tools/mcps, and align with what agents really want - instructions and scripts. but skills tell only part of the story, how to run the skills is the actual problem to solve for - which means you need a system for how to work in your codebase.
you should be able to start up a fresh CC/Codex/OpenClaw/etc session and ask "ok, whats next?" - if you dont get a good answer out of that, and quickly, your system is not aligned and coherent to get work done.
i think a lot of the systems out there like Gas Town and Superpowers are great, but i dont think its yet time to just adopt a system from someone else. things are moving too fast, you need to be agile, and understand the ins and outs of how you do work in a codebase.
a better mental model for your "do work" harness is the shadcn/ui route - you dont import a library, you add the atomic pieces that you will compose views from. they are yours to update and make part of the codebase - not some integration that you have to fight. this is where skills start to kinda get it right, but the framework around the skills ... you should own in your code. they should compound for YOU, not for the person building the framework. they understand that system in a different way than you ever will. your system should move in your direction.
the funny thing is, i think this is all noise. and we are not yet building around the right fundamental pieces.
the code should be an artifact of the spec. the code should be disposable, rebuildable at will by the next best model. the spec should be durable, yet modifiable. the code should realign around the spec.
i say this as someone with repos littered with either a /specs or /docs folder, who freaking knows the state of all of these. i assume they are "done" if im not actively working with them. but those specs and docs are constantly fighting against the code being reshaped, one feature or new spec at a time. there is no source of truth for "how should this work?" other than the code. which at this point, i dont think is the right way to go about things.
a spec breaks into 2 initial pieces: what and how.
the "what" defines your features. the "how" describes the tech stack and conventions - this is where skills should live and be activated from.
the reference for how to build forms should not be the first form that the agent decided to build in whatever way it chose. it should be a clearly defined form convention that points at your preferences/skills for how to build forms.
once we have the what and how, a system to do work is required, which means a system to break the spec down into "units of work" is also required. and this is the main battle ground everyone is fighting over right now, but it seems like focusing on how to define the what and how is more important, and will inform the work phase in a better way.
we should have swappable specs and tech stacks. you want your app in flask, ruby, node, astro, rust .. whatever, swap out the tech stack, the "how". we need to start aligning for the personalization layer, where people dont even want your app anymore, or your tech stack, they just want a reliable spec to feed into their agent, so it can get built out into their personal ecosystem.
bottom line: the code is not the source of truth. the spec is. the code is an artifact of the spec. specs should be durable, swappable, upgradable, and yours. it's time to refocus on the spec layer.
and the slope is accelerating, look to where we are going, not where we are.
xitter.cf/karpathy/status/200866…
3 biggest step changes in my AI journey: GPT-3.5, Sonnet 3.5, Opus 4.5
autocomplete → view composition → system composition
we're orchestrators now. the work shifted - planning, decisions, catching bad patterns early. typing isn't the job anymore.
smarter, faster, cheaper - all coming. but what we have works. you just have to ask the right questions.
Plan mode in cursor = very good
Now instead of begging the agent to make a plan and ask me questions, it's built in.
Describe feature, pull in initial files you know will be referenced and/or those that have the pattern that needs to be replicated.
Let it rip for a bit, then answer questions and provide feedback on the markdown plan file. Loop.
Once plan is ready, click build, let it rip again. At that point the feature is generally done, maybe a couple tweaks or a final cleanup prompt.
Using sonnet 4.5 thinking max (1m context).
EZ mode.
Join us as we talk AI 102: Running Open-Weight LLMs Locally
What are they? How to find them? How to run them? What hardware is needed?
About to start!
Life is full of choices 🔔
Never fade a chance to learn & skill up
Interacting with LLMs is the next most important tooling since you learned to read
Come hang and learn to improve the way you interact with the world, and yourself
xitter.cf/i/spaces/1kvJpylmdrMxE
We will talk through a lot of this in the upcoming community spaces tomorrow, Thursday at 6pm EST with @pepecoiners
I would love to hear your questions!
xitter.cf/pepecoiners/status/190…
Fantastic & easy to comprehend guide to using LLMs as success drivers in your personal & professional life 👇
Come join us Thursday 6pm EST on our Tech Talk Spaces for a Q&A with @getzerod 🐸🫡
You'll find the 🔔 link here: (xitter.cf/i/spaces/1jMKgkDEdXwKL)
Links
Perplexity @perplexity_ai
perplexity.ai/
Claude AI from Anthropic @AnthropicAI
claude.ai/
Super Whisper @superwhisper
superwhisper.com/
Cursor @cursor_ai
cursor.com/
ChatGPT from OpenAI @OpenAI
chatgpt.com/
Grok from X @grok
grok.com/
AI 101: TL;DR
Getting Started
- Recommended paid tools: Perplexity AI ($20/mo) for research, Claude AI ($20/mo) for writing, Cursor ($20/mo) for coding, Super Whisper ($8.49/mo) for voice-to-text prompting
- Paid services offer better privacy, features, and limits than free alternatives
- Total investment of ~$70/month pays for itself by saving just 1-2 hours of work
Everyday Use Cases
- Information organization: brain dumps, to-do lists, summarizing
- Writing assistance: emails, documents, content generation
- Research and learning: up-to-date information, concept explanations
- Analysis and decision-making: evaluating options from multiple perspectives
Effective Prompting
- Voice prompting (tools: Super Whisper, Wispr Flow) is faster than typing
- Structure prompts with: persona, goal, context, output format
- Use iterative prompting: start broad, refine through multiple interactions
Tool Strengths
- Perplexity: live web search, research, product comparisons
- Claude: document generation via Artifacts, image/document analysis
- Cursor: context-aware assistance with multiple AI interaction modes
Advanced Techniques
- Break complex tasks into manageable pieces
- Specify desired output formats for better results
- Use AI for personalized learning and skill development
Privacy Considerations
- Trade-offs between convenience and privacy
- Options from closed commercial models to locally-run open source
- Emerging technologies: Based AI and ZK LLMs for encrypted prompts
Getting Started Guide
- Begin with 1-2 tools and master them before expanding
- Practice voice prompting for brain dumps
- Develop personalized templates and workflows
- Aim to save 3-5 hours weekly within 30 days
Join us on the community spaces with your questions, 20 min! xitter.cf/pepecoiners/status/190…
Further details from the @getbasedai team here. xitter.cf/getbasedai/status/1905…
THE BASED THESIS
culture + tech > culture or tech
memes + utility > memes or utility
tech projects: serious, corporate, detached
meme projects: fun, but ultimately empty
BasedAI + Pepecoin play a different game
think bigger, play the long game
---
The Foundation:
decentralized compute network + ZK-LLM privacy layer
but that's just the beginning
Creatures are the gateway:
digital organisms that think and evolve 24/7
processing what matters to you while you touch grass
proactive agents, not just passive responders
They maintain balance across six dimensions:
- Emergence: developing novel properties
- Coherence: system stability and coordination
- Resilience: adapting to changes
- Intelligence: learning and decision-making
- Efficiency: optimal resource usage
- Integration: system connectivity
This isn't just prompt/response
This is true digital life
---
Brain Arena changes everything:
creatures are the apex of sentient AI agents
competing to generate the best plans and code
natural selection for digital intelligence
backed by real compute, real tech, real vision
1024 brains can exist on the network
342 can be active from day one
validators, miners, stakers working together
burn Pepecoin or stake 100k to join
---
This is for the people:
builders get a powerful compute network
everyone gets evolving digital creatures
privacy-first, trustless, decentralized
Creatures are just the first application
builders will create more dApps
normal users become digital architects
everyone can be part of this revolution
The future is based:
creatures collaborating with creatures
knowledge and capabilities compound
network effects of digital intelligence
moving towards collective enhancement
We're building something bigger than ourselves
but we're having fun doing it
be true, be real, be based
say what you mean, do what you say
that's what being based means
serious things can be fun
life is too short to be serious all the time
but the tech must be real
decentralized, private, trustless, and agentic
this is where we naturally head
---
BasedAI + Pepecoin is for the people
focus on the user
build utility
make it fun
memes alone are capped
think bigger anon
Official links:
Pepecoin @pepecoins
pepecoin.io/
coinmarketcap.com/currencies…
0xA9E8aCf069C58aEc8825542845Fd754e41a9489A
BasedAI @getbasedai
getbased.ai/
coinmarketcap.com/currencies…
0x44971ABF0251958492FeE97dA3e5C5adA88B9185
Based creature running.
First thought parsed ... this is the beginning of something totally different.
Focus up, frens, @getbasedai is cooking.
Gone through about $2 in credits over the last 30 min
The base model that is being used is `x-ai/grok-beta`
- $5 per million tokens in, $15 per million tokens out
- line 53 in `src/api/openrouter.rs`
- openrouter.ai/x-ai/grok-beta
No UI on this version
Code is in rust
The nature of the burn has been fully revealed.
9/30/2024 4:20pm EDT
Burn our beloved @pepecoins to receive Brain Credits. 1000 Brain Credits = 1 Brain, of which there can only ever be 1024 of. You can also stake 100k pepecoin for 1 whole Brain, but we will be discussing the burn in this post.
What are we in for? Let's go through all the steps that need to happen to get through the burn.
- 24 hours before the burn a snapshot will be taken of staking rewards
- Sometime around 9/29/2024 4:20 EDT the grace period will open, and we will be able to unstake our pepecoin and basedai (be sure to get full, final word from the team before unstaking)
- When you unstake, everything will be unstaked, this is how the contract works
- Our rewards are locked in at this point, we will be able to claim them when mainnet goes live, we don't know how long it will be after the burn until mainnet is released
- This leaves us with about 24 hours to move our pepecoin around, figure out what wallet we will burn from, and have your ETH ready to be used as gas. This will be our last chance to finalize our plan. How much are you going to burn? How much gas are you willing to put up?
- The burn will be first-come-first-serve
- The burn portal and contract will open up at 9/30/2024 4:20pm EDT
At this point, if you have not watched the recent video from @getbasedai, please do so. This is a walk through of all the screens and options we will be presented with. You would be wise to study hard while watching.
Let's break the video down into pieces.
- Go to the burn portal website - we don't have a link yet
- Connect your wallet - I'm assuming this will be the same tech as used on the pepecoin.io website, probably worth running through connecting your wallet to that site from a private window to ensure this is still a smooth process for you
- From the "Brain Management" screen, click "Acquire Brains"
- Then from "Acquire Brains via" click "Burning"
- Now we have our burn options to fill out ...
- Click into the input and set your "Pepecoin to Burn" amount (you know how many you're going to burn, right? kek)
- You will see an estimated amount of Brain Credits you will receive from this amount of pepecoin (but if you are burning in the first seconds, don't assume this is the amount you will receive)
- Now we have an optional field "SET MAX PRICE PER 1k CREDITS" ... heh, think about this one for a moment
This is the field to set if you want to bail on the transaction (and NOT burn your pepecoin) if Credits get too expensive for what you want. 1k Credits is 1 Brain. What's a Brain worth to you? Is there a point where you want to say "hey, stop, don't burn my pepecoin, just give them back"? We have no idea how hot this is going to be in the first moments, but we are expecting fireworks. If you simply want the best rate you can get, no matter where you end up in the burn order, you do not want to set this field. The rate will only go up. Think hard about this one now, and set the field if this is your decision.
- Ok, so we've set our pepecoin to burn and optionally set our bail out rate ...
- Transaction #1: Click "APPROVE PEPECOIN" - note, this is NOT burning, this is approving your wallet to be allowed to spend your pepecoin for the burn
- Approve the spending limit in your wallet (quickly I might add), set your gas accordingly
- Once this transaction has cleared, you will see a button for "BURN" appear on the website
- Transaction #2: Click "BURN" - this is the big one, this is sending your pepecoin into the contract to be executed in the order received
- Approve the burn transaction in your wallet, set your gas accordingly
- Now fucking pray to Kek that your expectations meet the reality that is about to be shown to you ... those pepecoin are gone, and you will get Brian Credits in return
If you were lucky enough to get 1000 or more Brain Credits, well, you will be able to now mint 1000 of those into a Brain NFT.
We’re very excited to announce that the date and time for the Burn for Brains event is set to take place on September 30, 2024 at 4:20pm EDT. The final contracts are in the process of being prepared for third-party audit. Pending successful completion of the audit, the burn will commence at the scheduled time, and Brain redemption will be available immediately.
In the meantime, we have created a video that provides a step-by-step walkthrough of the Burn, Stake, and Redemption Portal, covering all the key actions you can take:
Burning PepeCoin for Brain Credits
Redeeming a Brain for full ownership
Staking PepeCoin for a Brain rental
Partial Pooling for shared Brain ownership
This video will help guide you through each option available in the portal, so you’ll be ready to participate in this exciting opportunity if desired.
Thank you for your patience and enthusiasm as we get closer to this historic event. Stay tuned for more updates!
What am I doing ... HAH, wouldn't you like to know?
Like many, I have a range of what I'm willing to burn, a high and a low number set. The final burn amount will depend on the price at time of burn, and the overall vibe. I'm leaning somewhere in the middle of my range. No, I'm not posting my amounts here, but I do intend to get myself at least 1 Brain.
Think real hard on what you are willing to put up on unseen tech. There is only one option for me - burn. This is why I'm here. This is what hooked me on the project.
I build in AI daily, but mainly using centralized models. However, the rate at which open-weights models are improving is faster than imagined. Seems like weekly we are consistently topping the charts with what is possible. My initial goals are to move as much of what I do daily onto based for my real-world clients. I have so many ideas I want to test once I get my hands on the tech.
I believe in a decentralized, private, trustless AI future where everyone has access to intelligence.
BasedAI is the only project I've seen that checks all these boxes. It offers an open system where:
- Builders are empowered to solve for their specific use-cases
- Validators and miners can contribute hardware to power the decentralized network
- End users are equipped with private, verified intelligence
The future is $BASED.
Well, they gave themselves an out and used it. Burn delayed.
New burn date
10/2/2024 4:20pm EDT
Which means we should be able to unstake sometime around 4:20pm EDT 10/1.
Le sigh.
xitter.cf/getbasedai/status/1840…
PepeCoin and BasedAI Community,
We want to take a moment to update you on the upcoming Burn for Brains event. We are updating the schedule by 48 hours to accommodate additional audit time.
The Burn is now planned to commence at 4:20pm EDT on the 2nd of October.
Within this time an announcement will also be made regarding the grace period snapshot - so please wait for official confirmation of the snapshot before unstaking.
Burn baby, burn. 🔥🐸🔥