SUMM

Itamar Friedman introduces himself as CEO and co-founder of Codto, announcing the soft launch of their new CLI tool
He lays out the plan to discuss the evolution of developer tools, particularly focusing on CLI as a future interface for confident coding
Describes aims to introduce "vibe coding with confidence" and why current tools have yet to be true game changers

Outlines three generations: Gen 1.0 (autocomplete), Gen 2.0 (chat-based, agentic coding), and the anticipated Gen 3.0 (command-driven agent workflow)
Gen 1.0: Autocomplete offered productivity increase but wasn't transformative
Gen 2.0: Chat agents allowed more code generation, particularly empowering junior developers, but created code review challenges and quality concerns
Gen 3.0: The forthcoming shift is command-line control over agents, enabling comprehensive end-to-end workflows in the software development lifecycle (SDLC)

Differentiates "noobs" (new developers) and "enterprise" (experienced, complex environments)
For less experienced users, end-to-end can mean turn-key software generation
For enterprise, true value requires AI support for versatile, reliable workflows across the SDLC—code generation, testing, quality review, refactoring, etc.
Current tools offer siloed AI capabilities; the next game changer will holistically integrate workflows, especially red-team (review, quality, testing) processes

References Andrej Karpathy's discussion of vibe coding, initially celebrated for its productivity, then questioned for code quality and context loss
Emphasizes the need for context-rich, workflow-driven code production, especially for "code I actually and professionally care about"
Quality assurance, context gathering, and workflow connection must become integral, not manual, steps

AI tools perform well in greenfield (brand new) projects, but struggle with large, legacy codebases due to maintainability, testing, and review demands
Codto aims to address these challenges holistically using a multi-agent architecture for research, code review, and best practices integration

Brief overview of Codto agents: deep research agent, code review agent (codemerge), and context/best practice gathering (codaware)
The newly launched CLI tool is designed to integrate workflows and agents directly into developers' processes, with a soft launch announced during the talk
The CLI enables creation and customization of agents and workflows that automate research, code generation, coverage improvement, and code review tasks

Highlights the flexibility and workflow orchestration possible with CLI tools, citing an example from another presenter's session
CLI allows developers to run, compose, and automate agents and workflows, including piping output between tools in customized ways
CLI empowers both sequential and parallel (A2A, agent-to-agent) workflows that surpass what is feasible with IDE plugins

CLI facilitates chaining multiple tasks (e.g., code generation, coverage improvement, review) in one streamlined process
Introduces the idea of agent-to-agent (A2A) communication: agents operating in parallel, exchanging information, and coordinating actions
Although not yet widespread, A2A is seen as the next evolution for developer productivity and reliability

Codto's CLI tool allows enforcement of organizational best practices, coverage criteria, and integration with pre-/post-commit hooks for confidence-inspiring development
CLI facilitates automatic generation of task-specific interfaces, reducing dependency on traditional IDEs
Friedman predicts a near future (2025–2026) where swarms of specialized agents, via flexible interfaces like CLI, will dominate advanced software development workflows

Vibe Coding with Confidence — Itamar Friedman, Qodo