Moeed retweeted
In the next version of Bun
Bun.serve() gets a simple builtin router. It supports params.
![import { serve, sql } from "bun"; import App from "./myReactSPA.html"; serve({ routes: { "/*": App, "/api/users/:id": async (req) => { const id = req.params.id; const [user] = await sql`SELECT * FROM users WHERE id = ${id} LIMIT 1`; return Response.json(user); }, "/api/users": { POST: async (req) => { // ... }, }, }, fetch: () => new Response("Not Found", { status: 404 }), });](https://pbs.twimg.com/media/Gj05utSagAAV14Q.jpg?name=small&format=webp)
ALT import { serve, sql } from "bun"; import App from "./myReactSPA.html"; serve({ routes: { "/*": App, "/api/users/:id": async (req) => { const id = req.params.id; const [user] = await sql`SELECT * FROM users WHERE id = ${id} LIMIT 1`; return Response.json(user); }, "/api/users": { POST: async (req) => { // ... }, }, }, fetch: () => new Response("Not Found", { status: 404 }), });
Moeed retweeted
Agents to Maximize Human Efficacy
How can we commoditize the effectiveness of a billion dollar enterprise for individuals?
When the tractor was first introduced, many feared it would steal jobs, pollute the air, and upend society. Yet, rather than causing harm, it became an indispensable tool—revolutionizing farming, driving economies, and reshaping entire societies. By boosting productivity, it ultimately improved the well-being of billions—all from the unassuming tractor.
A tractor isn’t valuable, simply because it can pull farming equipment. Its true impact lies in the surplus productivity it generates. By increasing food security, it enables people to focus on higher pursuits—self-actualization, innovation, and enhancing the productivity of others. The benefits are more holistic than farming.
Humanity's most effective tool today is the computer. By moving information from A to Z you can automate tasks, foster human connections and access humanities collective intelligence. Its effectiveness is only bound by the software it runs.
It begs the question, what is the software that will unlock humanity’s maximum effectiveness.
Agents are autonomous machines that think, learn, and act. They navigate the complex web of data we live in, making sense of the chaos to take meaningful actions. From emails and documents to social media and financial records, they analyse patterns and extract insights to make real-time decisions. Whether scheduling meetings, optimizing workflows, or executing transactions, agents convert raw information into intelligent action, amplifying human capability and accelerating progress.
By providing a platform where agents can rapidly develop and self-optimize we propose a system that facilitates an ecosystem of AI agents. We bring these agents together under a super agent—capable of solving any problem. Just as Google indexes all information, the super agent indexes the space of all actions. While the internet democratized knowledge, the super agent is democratizing productivity.
Internet v1.0 - Made information rapidly available via Facebook, Google and Wikipedia
Internet v2.0 - Made actions rapidly available, optimizing for efficacy
The Path to Super AI Agents
Conceptualizing a super AI agent which:
- Processes and analyses your personal data, along with the vast amounts of global information sourced from the internet.
- Consider all possible actions that can be taken on your data, via a global action search index.
- Measures metrics generated by your actions, to self optimize for greater results.
Employing an incremental approach to actualizing this super AI agent; by creating a system of generating and optimizing specialized agents. The system itself is simple:
- Store actions as Javascript code.
- Make the actions searchable.
- Inject actions into traditional tool-use LLM agents.
- Build a system of maintaining and organizing the action-space.
Here we outline the key, guiding principles of this system.
The Action Space
Traditional large language model (LLM) agents solve problems by executing tasks in a sequential loop of action and observation. While this process is effective, it becomes inefficient, particularly for complex tasks that involve long chains of actions. The Action Space offers a solution by compressing these multi-step processes into single function calls. Here, the LLM is used for higher-order thinking and planning, injecting decision-making exactly where needed in the tree of agent and function calls. This hybrid model merges the efficiency of code execution with the adaptability of AI-driven decision-making, streamlining the problem-solving process.
Actions in the system are generated on the fly—whether by a user or an agent—when creating an agentic app or executing a task. Once created, these actions are indexed in a searchable database for easy retrieval and reuse.
To further optimize the system, we focus on organizing actions for maximum reuse and minimal redundancy. Traditional software development often suffers from fragmented code and extensive duplication. To address this, we’ve developed a unified repository with exactly one solution to every problem. We use a log-structured merge tree (LSM tree) to collect high-entropy functions, identify patterns, and recognize dependencies. Reusable components are then merged into the larger structure, ensuring that every action is placed in the right context.
This organized action-space enables continuous improvement. Actions can be refined, expanded, and recombined, fostering a dynamic and adaptable system. The action space is not centralized, but rather collaborative, with a community of human and AI maintainers ensuring the quality and relevance of each action.
By addressing redundancy and fragmentation through a well-structured, global action space, we create a foundation for efficiency, scalability, and ongoing optimization—driving progress at the intersection of code and AI.
Metric Optimization
Driving the evolution of the action space is the optimization of metrics. By systematically measuring the outcomes of each action, analyzing what drives changes in performance and feed back these insights into the system, we create a self improving system focused on maximizing efficacy.
Conclusion
The Super Agent serves as the foundation where human ingenuity is augmented rather than replaced.
By transforming raw information into intelligent, goal-oriented action, we empower individuals to focus on higher-order thinking, creativity, and innovation.
With a decentralized, community-driven approach to AI development, we ensure that the benefits of this technology remain open, transparent, and accessible to all.
In the same way the tractor revolutionized agriculture and the internet revolutionized information, agents are set to revolutionize productivity—ushering in a new era where AI agents work alongside us to achieve the greatest version of human potential.
Moeed retweeted
valentines day is cool and all, but is it distracting you from building an agentic bootstrapped 100m arr ai b2b vertical saas startup?
Moeed retweeted
Central to our mission is using agents to maximise human potential. Check out how we plan to do so in our white paper below:
Moeed retweeted
The humble tractor skyrocketed human food security.
It redefined economies and civilisation. It's value was more holistic than it's ability to tug farming equipment.
AI agents will skyrocket human efficiency, paving a path to self actualisation.
A founder's letter on how technology fucks us up, and what we are doing about it:
linkedin.com/feed/update/urn…
159
Interested in how LLMs like Chat-GPT work under the hood? Check out this open source model I cooked up this weekend which uses the same tech as Chat-GPT but on a smaller scale.
github.com/MoeedDar/gpt-rs
I suck at content production and I'm not afraid to hide it ;) .
If you're worried about the quality of your videos, put something out there anyways. The point of @_buildspace is to iterate and improve. You cannot improve until you're comfortable to fail. @_nightsweekends
This quoted post is unavailable.
1
9
321








