All activity
Andrii Shavelleft a comment
Really cool to see this direction. The portability issue between Claude Code, Cursor, Copilot, etc. is something I’ve run into as well — each agent having its own “skill dialect” makes it hard to build consistent workflows. I’m working on a different layer of the stack (Iceberg Framework), where the focus is on deterministic execution and validation for LLM‑powered systems. What you’re doing...
SkillkitThe package manager for AI agent skills
Iceberg Framework introduces a rules layer for AI — something existing tools don’t provide.
Instead of generating code directly from prompts, Iceberg forces models to follow a structured process: rules → documents → code.
This eliminates behavioral drift, reduces variance, and keeps systems maintainable over time.
It works behind the scenes, adding predictable, spec‑driven behavior to any model without changing your workflow.
Iceberg FrameworkRules for AI.
Andrii Shavelstarted a discussion
What problems does Iceberg Framework solve for you?
Iceberg Framework was created to make LLM‑driven workflows predictable, validated, and safe to run in real systems. But every team and developer faces different challenges when working with AI. I’m curious to hear from the community: What problems do you struggle with when building agentic workflows? Where do LLMs behave unpredictably in your projects? Do you need validation, reproducibility,...
Andrii Shavelleft a comment
Hi everyone 👋 I’m Andrii, the solo maker behind Iceberg Framework. What is Iceberg Framework? Iceberg is a deterministic execution layer for LLM‑powered systems and agents. It makes AI workflows predictable, validated, and production‑ready — even when using non‑deterministic models like GPT or Claude. Most AI systems break because LLMs hallucinate, produce inconsistent outputs, or silently...
Iceberg FrameworkRules for AI.
Andrii Shavelleft a comment
Congrats on the launch — love the idea of giving coding agents a controlled execution environment. One question from a standards/consistency perspective: how granular can the sandbox rules be? Do you allow defining reusable rule sets or patterns for file access, so the agent behaves consistently across different projects?
MultituiSandbox claude code, codex, or any TUI on macOS
Andrii Shavelleft a comment
Congrats on the release — impressive work. I’m curious from a standards/consistency perspective: when SERA adapts to a new repo, how do you ensure it follows stable patterns instead of drifting between different coding styles? Is there any way to define explicit rules or constraints the model must follow during generation?
SERAFast, accessible coding agents that adapt to any repo
Andrii Shavelleft a comment
You mention vibe‑coding as a core part of the workflow. How do you prevent the AI from drifting or producing inconsistent changes across iterations?
CreateOSBuild and deploy apps from any AI coding tool, in one place
Andrii Shavelleft a comment
Congrats on the launch — really interesting approach to agent‑driven storefronts. I’m curious about the technical side: are your AI agents based on a trained internal model, or are they role‑specialized through prompting? In other words, does each agent have its own “persona” defined by prompts, or did you build something more structured behind the scenes? Would love to understand how you...
Genstore.aiTest, iterate, and launch an agentic storefront in minutes
Andrii Shavelleft a comment
Congrats on the launch. Curious — how do you handle unpredictable behavior in AI‑driven components of the system? That’s where I see most real‑world failures.
ArchimystAI-powered platform for designing system architecture