PinnedIvan·Apr 19MLX vs CoreML on Apple Silicon: A Practical Guide to Picking the Right Backend — and Why You Should…You’re building a speech feature on Apple Silicon. You have two backends. Pick the wrong one and your voice activity detection runs 50x…A response icon1A response icon1
PinnedIvan·Feb 12Qwen3-ASR Swift: On-Device ASR + TTS for Apple Silicon — Architecture and BenchmarksQwen3’s speech models — both ASR and TTS — are impressive, but when I looked for a Swift implementation to run them on Apple Silicon, there…A response icon1A response icon1
PinnedIvan·Feb 23NVIDIA PersonaPlex 7B on Apple Silicon: Full-Duplex Speech-to-Speech in Native Swift with MLXWhat if you could talk to your laptop and it talked back — not through a three-step pipeline of transcribe-think-synthesize, but as a…A response icon3A response icon3
Ivan·Mar 20We Beat Whisper Large v3 with a 600M Model Running Entirely on Your MacWhat if the best speech recognition in the world didn’t live in a data center? What if it ran faster than real-time, on your laptop, with…A response icon1A response icon1
Ivan·Feb 26Speaker Diarization and Voice Activity Detection on Apple Silicon: Native Swift with MLXThe speech-swift library keeps growing. First came ASR and TTS — turning speech into text and text into speech. Then PersonaPlex —…
Ivan·Feb 7Context Engineering for Agentic Hybrid Applications: Why Code Agents Fail and How to Fix ThemMemory layers, tool compression, MCP, trajectory pruning, and meta-agent CI for scalable code agents in 2026.
Ivan·Dec 4, 2025Claude Code vs OpenAI Codex: Agentic Planner vs Shell‑First SurgeonBoth Claude Code and OpenAI Codex are built on the same backbone: a single-agent event loop that repeatedly thinks, calls tools, inspects…
Ivan·Nov 22, 2025Beating Qwen3 LoRA with a Tiny PyTorch Encoder on the Large‑Scale Product CorpusIn this post I revisit product category classification on the Large Scale Product Corpus V2020 dataset with a fresh approach. Instead of…
Ivan·Oct 26, 2025Inside the Modern Search Engine: The Complete Pipeline — How It Ranks Results & Learns from…Modern web search isn’t magic; it’s a massive engineering feat. Crawlers scour billions of pages, hybrid indexes organise that data, query…A response icon1A response icon1
Ivan·Aug 31, 2025Practical guide: fine-tuning Qwen3 with LoRA. KL-anchored SFT and β-tuned DPOA practical guide to KL-anchored SFT and β-tuned DPOA response icon2A response icon2
