An Agent Framework Hiding Inside a Session: Dynamic Profiles in Foundation Models
Dynamic Profiles turn one LanguageModelSession into a declarative agent runtime: per-phase models, history transforms, and two orchestration patterns.
Ivan Magda
Writing about AI Engineering and Software Development.
Latest posts
A Server LLM With No API Keys: Private Cloud Compute in the Foundation Models Framework
Private Cloud Compute gives apps a 32K-context server model with reasoning and no API keys and makes the per-user daily quota a UI design problem.
Foundation Models, Year Two: From On-Device API to General LLM Runtime
WWDC26 turned Foundation Models into Swift's general LLM client: a rebuilt on-device model, a server model with no API keys, and the first real agent primitives.
Building a Coding Agent in Swift, Part 0: Bootstrapping the Project
Setting up a two-target SPM package, strict concurrency, and AsyncHTTPClient — the foundation for a CLI agent that never needs restructuring.
Five Things I Learned About Making AI Coding Agents Actually Work
The biggest gains in AI agent performance come from the system around the model, not the model itself.
I Built a Coding Agent From Scratch in Swift. Here's What I Missed Until I Read Koog.
After nine posts building a Claude Code-style agent by hand in Swift, I opened up JetBrains Koog framework and realized how much of what I'd hand-rolled was already someone else's problem. Here's what the framework gave me for free, and what it changed about how I think about agents.
Fixing the 40k CLAUDE.md Warning in a Monorepo
How we restructured Claude Code's CLAUDE.md files in a monorepo to cut 73% of always-loaded context without losing a single fact.
Building a Coding Agent in Swift, Part 8: Background Tasks
Swift actors for background execution — the one type that truly needs an actor, with notification injection before each API call.
Building a Coding Agent in Swift, Part 7: Task System
A file-based task DAG with dependency resolution — durable planning that survives context compaction.
Building a Coding Agent in Swift, Part 6: Context Compaction
A three-layer compression strategy — micro-compaction, auto-compaction, and manual compaction — for infinite conversations.
Building a Coding Agent in Swift, Part 5: Skill Loading
Two-layer knowledge injection — cheap awareness in the system prompt, full skill bodies loaded on demand via tool results.
Building a Coding Agent in Swift, Part 4: Subagents
Context isolation through recursive agent loops — spawning child agents with restricted tool sets for focused subtasks.
Building a Coding Agent in Swift, Part 3: Self-Managed Task Tracking
A Codable todo system with validation constraints and nag reminders — giving the agent a structured notepad it writes for itself.
Building a Coding Agent in Swift, Part 2: Tool Dispatch
Dictionary-based tool dispatch, path sandboxing, and the read/write/edit file tools that let the agent work with code.
Building a Coding Agent in Swift, Part 1: The Agent Loop
The agent loop that drives everything — how a single stopReason check turns an API client into an autonomous agent.