To avoid the rapid depreciation of hard-coded systems as LLMs improve, Blitzy's architecture is dynamic. Agents are generated just-in-time, with prompts written and tools selected by other agents based on the latest model capabilities and the specific task requirements.
Related Insights
Instead of switching between ChatGPT, Claude, and others, a multi-agent workflow lets users prompt once to receive and compare outputs from several LLMs simultaneously. This consolidates the AI user experience, saving time and eliminating 'LLM ping pong' to find the best response.
True Agentic AI isn't a single, all-powerful bot. It's an orchestrated system of multiple, specialized agents, each performing a single task (e.g., qualifying, booking, analyzing). This 'division of labor,' mirroring software engineering principles, creates a more robust, scalable, and manageable automation pipeline.
An AI agent uses an LLM with tools, giving it agency to decide its next action. In contrast, a workflow is a predefined, deterministic path where the LLM's actions are forced. Most production AI systems are actually workflows, not true agents.
To avoid confusing agents with contradictory goals, Tasklet plans to shift from pre-generated, static instructions to dynamically generating them just-in-time for each task run. This ensures the agent always operates on the most current user feedback, preventing errors from conflicting historical directives.
The true building block of an AI feature is the "agent"—a combination of the model, system prompts, tool descriptions, and feedback loops. Swapping an LLM is not a simple drop-in replacement; it breaks the agent's behavior and requires re-engineering the entire system around it.
In this software paradigm, user actions (like button clicks) trigger prompts to a core AI agent rather than executing pre-written code. The application's behavior is emergent and flexible, defined by the agent's capabilities, not rigid, hard-coded rules.
The best UI for an AI tool is a direct function of the underlying model's power. A more capable model unlocks more autonomous 'form factors.' For example, the sudden rise of CLI agents was only possible once models like Claude 3 became capable enough to reliably handle multi-step tasks.
Early agent development used simple frameworks ("scaffolds") to structure model interactions. As LLMs grew more capable, the industry moved to "harnesses"—more opinionated, "batteries-included" systems that provide default tools (like planning and file systems) and handle complex tasks like context compaction automatically.
The pace of AI model improvement is faster than the ability to ship specific tools. By creating lower-level, generalizable tools, developers build a system that automatically becomes more powerful and adaptable as the underlying AI gets smarter, without requiring re-engineering.
Replit's leap in AI agent autonomy isn't from a single superior model, but from orchestrating multiple specialized agents using models from various providers. This multi-agent approach creates a different, faster scaling paradigm for task completion compared to single-model evaluations, suggesting a new direction for agent research.
