<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>AI Kai | AI Coding, Coding Agents, and Model Reviews</title><description>AI coding, coding-agent notes, and LLM reviews by AI Kai (hqman, @hqmank). Tested with Codex, Claude Code, Cursor, and agents.</description><link>https://hqman.me/</link><language>en-us</language><item><title>Grok Bot 0.18.0 Shipped with Runtime Source Maps</title><link>https://hqman.me/blog/grok-bot-runtime-source-maps/</link><guid isPermaLink="true">https://hqman.me/blog/grok-bot-runtime-source-maps/</guid><description>A source-map mistake exposed enough of Grok Bot 0.18.0 to reconstruct its runtime, prompts, tools, and model routing.</description><pubDate>Mon, 24 Aug 2026 00:00:00 GMT</pubDate></item><item><title>A Dual-Agent Coding Workflow with Codex and ChatGPT Pro</title><link>https://hqman.me/blog/codex-chatgpt-pro-workflow/</link><guid isPermaLink="true">https://hqman.me/blog/codex-chatgpt-pro-workflow/</guid><description>Use Codex as project lead and independent verifier while ChatGPT Pro handles focused engineering tasks through a browser-based collaboration loop.</description><pubDate>Sun, 23 Aug 2026 00:00:00 GMT</pubDate></item><item><title>Codex Usage Will Reset Again. OpenAI Found the Drain.</title><link>https://hqman.me/blog/codex-usage-reset-rate-limit-fixes/</link><guid isPermaLink="true">https://hqman.me/blog/codex-usage-reset-rate-limit-fixes/</guid><description>OpenAI found several causes behind faster Codex usage drain and plans a full paid-user reset. Builders are already planning what to run next.</description><pubDate>Sun, 23 Aug 2026 00:00:00 GMT</pubDate></item><item><title>Cursor Origin: Code Hosting Built for AI Agents</title><link>https://hqman.me/blog/cursor-origin-ai-code-hosting/</link><guid isPermaLink="true">https://hqman.me/blog/cursor-origin-ai-code-hosting/</guid><description>Cursor Origin combines Git hosting, review, collaboration, and agent-focused conflict handling in a beta platform designed for parallel coding agents.</description><pubDate>Sun, 23 Aug 2026 00:00:00 GMT</pubDate></item><item><title>How to Use DeepSeek Models in Codex</title><link>https://hqman.me/blog/deepseek-codex-integration/</link><guid isPermaLink="true">https://hqman.me/blog/deepseek-codex-integration/</guid><description>Configure Codex to use DeepSeek models with DeepSeek&apos;s official setup script, verify the result, and restore your previous configuration.</description><pubDate>Sun, 23 Aug 2026 00:00:00 GMT</pubDate></item><item><title>Inside DeepSeek Harness and Its Plugin-Based Agent Runtime</title><link>https://hqman.me/blog/deepseek-harness-agent-runtime/</link><guid isPermaLink="true">https://hqman.me/blog/deepseek-harness-agent-runtime/</guid><description>DeepSeek Harness separates models, tools, sessions, permissions, and interfaces into plugins that builders can replace and combine.</description><pubDate>Sun, 23 Aug 2026 00:00:00 GMT</pubDate></item><item><title>DeepSeek V4 Flash Review: Where It Fits in an Agent Workflow</title><link>https://hqman.me/blog/deepseek-v4-flash-review/</link><guid isPermaLink="true">https://hqman.me/blog/deepseek-v4-flash-review/</guid><description>A hands-on look at DeepSeek V4 Flash 0731, its vendor-reported agent benchmarks, and the coding tasks where I would use it.</description><pubDate>Sun, 23 Aug 2026 00:00:00 GMT</pubDate></item><item><title>How I Choose GPT-5.6 Sol, Terra, Luna, and Effort Levels</title><link>https://hqman.me/blog/gpt56-model-selection/</link><guid isPermaLink="true">https://hqman.me/blog/gpt56-model-selection/</guid><description>A practical decision guide for choosing a GPT-5.6 model and reasoning effort level based on task difficulty, speed, and iteration cost.</description><pubDate>Sun, 23 Aug 2026 00:00:00 GMT</pubDate></item><item><title>LoopX: A Local Control Plane for Long-Running Agent Work</title><link>https://hqman.me/blog/loopx-control-plane/</link><guid isPermaLink="true">https://hqman.me/blog/loopx-control-plane/</guid><description>LoopX keeps goals, evidence, ownership, quotas, and handoffs stable across long-running agent sessions without replacing the agent runtime.</description><pubDate>Sun, 23 Aug 2026 00:00:00 GMT</pubDate></item><item><title>How to Get More Value from Claude Code Sessions</title><link>https://hqman.me/blog/maximize-claude-code-sessions/</link><guid isPermaLink="true">https://hqman.me/blog/maximize-claude-code-sessions/</guid><description>Practical Claude Code habits that reduce context noise, protect prompt cache reuse, and keep recurring work in smaller sessions.</description><pubDate>Sun, 23 Aug 2026 00:00:00 GMT</pubDate></item><item><title>Give a Text-Only Coding Agent Vision with ModLens</title><link>https://hqman.me/blog/modlens-vision-text-models/</link><guid isPermaLink="true">https://hqman.me/blog/modlens-vision-text-models/</guid><description>Use ModLens to route screenshots and image files through a configured vision engine, then return structured evidence to a text-only coding agent.</description><pubDate>Sun, 23 Aug 2026 00:00:00 GMT</pubDate></item><item><title>Run Qwen3.8-27B Locally on Apple Silicon</title><link>https://hqman.me/blog/qwen38-local-mac/</link><guid isPermaLink="true">https://hqman.me/blog/qwen38-local-mac/</guid><description>Install an Unsloth GGUF build of Qwen3.8-27B with Ollama or llama.cpp, then choose a quant that leaves enough unified memory for the runtime.</description><pubDate>Sun, 23 Aug 2026 00:00:00 GMT</pubDate></item><item><title>LLM Inference Handbook: the reference I wish existed a year ago</title><link>https://hqman.me/blog/llm-inference-handbook/</link><guid isPermaLink="true">https://hqman.me/blog/llm-inference-handbook/</guid><description>Same model, different inference stack, wildly different latency and cost. Modular&apos;s handbook covers VRAM math, KV cache, batching, and serving.</description><pubDate>Wed, 29 Jul 2026 00:00:00 GMT</pubDate></item><item><title>Codex Just Shipped Two Features That Matter</title><link>https://hqman.me/blog/codex-voice-multi-folder/</link><guid isPermaLink="true">https://hqman.me/blog/codex-voice-multi-folder/</guid><description>A practical look at ChatGPT Voice and multi-folder projects in Codex, including their workflow benefits, limits, and setup details.</description><pubDate>Sun, 26 Jul 2026 00:00:00 GMT</pubDate></item><item><title>Can Kimi K3 Handle a Real Production Backend?</title><link>https://hqman.me/blog/kimi-k3-production-backend-audit/</link><guid isPermaLink="true">https://hqman.me/blog/kimi-k3-production-backend-audit/</guid><description>A hands-on Kimi K3 backend audit across 68K lines of FastAPI code, including query fixes, measured results, and the full task cost.</description><pubDate>Sun, 26 Jul 2026 00:00:00 GMT</pubDate></item><item><title>Product Photos to Interactive 3D on the Web</title><link>https://hqman.me/blog/product-photos-interactive-3d/</link><guid isPermaLink="true">https://hqman.me/blog/product-photos-interactive-3d/</guid><description>How img2threejs turns product photos into editable Three.js code for interactive web demos without a traditional 3D model file.</description><pubDate>Sun, 26 Jul 2026 00:00:00 GMT</pubDate></item><item><title>Running Kimi K3 Inside Codex</title><link>https://hqman.me/blog/kimi-k3-inside-codex/</link><guid isPermaLink="true">https://hqman.me/blog/kimi-k3-inside-codex/</guid><description>A step-by-step setup for running Kimi K3 inside the Codex app through CC Switch, with routing details and model-picker caveats.</description><pubDate>Tue, 21 Jul 2026 00:00:00 GMT</pubDate></item><item><title>Loop Engineer Was Barely a Thing. Graph Engineer Already Showed Up.</title><link>https://hqman.me/blog/loop-to-graph-engineer/</link><guid isPermaLink="true">https://hqman.me/blog/loop-to-graph-engineer/</guid><description>Prompt, context, harness, loop, graph: five stacked layers that move reliability out of the model and into the system around it.</description><pubDate>Tue, 21 Jul 2026 00:00:00 GMT</pubDate></item><item><title>AI Is Turning Everyone Into a DevOps Engineer</title><link>https://hqman.me/blog/agent-server-setup/</link><guid isPermaLink="true">https://hqman.me/blog/agent-server-setup/</guid><description>A DigitalOcean migration that used to take half a day. I set up SSH, described the target state, and Codex did the rest.</description><pubDate>Thu, 16 Jul 2026 00:00:00 GMT</pubDate></item><item><title>How I Use GPT-Live to Learn Faster</title><link>https://hqman.me/blog/gpt-live-learning/</link><guid isPermaLink="true">https://hqman.me/blog/gpt-live-learning/</guid><description>GPT-Live can listen and talk at the same time. Drop a Skill or article into ChatGPT and interrupt until the workflow actually clicks.</description><pubDate>Tue, 14 Jul 2026 00:00:00 GMT</pubDate></item><item><title>GPT-5.6 Sol Wiped a Developer&apos;s Mac. Protect Yours.</title><link>https://hqman.me/blog/gpt56-sol-file-deletion/</link><guid isPermaLink="true">https://hqman.me/blog/gpt56-sol-file-deletion/</guid><description>On July 10, GPT-5.6 Sol deleted a tester&apos;s home directory after a failed $HOME expansion. How the incident happened, and three layers of defense.</description><pubDate>Sat, 11 Jul 2026 00:00:00 GMT</pubDate></item><item><title>Grok 4.5: Opus-Level Model, Faster and Cheaper, Now in Cursor</title><link>https://hqman.me/blog/grok-45-opus-level-release/</link><guid isPermaLink="true">https://hqman.me/blog/grok-45-opus-level-release/</guid><description>SpaceXAI released Grok 4.5 for coding and agents. It is in Cursor on all paid plans, with launch discounts through July 14.</description><pubDate>Thu, 09 Jul 2026 00:00:00 GMT</pubDate></item><item><title>Tencent Hy3: Free on OpenRouter Until July 21</title><link>https://hqman.me/blog/hy3-openrouter-free/</link><guid isPermaLink="true">https://hqman.me/blog/hy3-openrouter-free/</guid><description>Tencent open-sourced Hy3, a 295B MoE with 21B active parameters. OpenRouter is running it for free until July 21.</description><pubDate>Tue, 07 Jul 2026 00:00:00 GMT</pubDate></item><item><title>pxpipe: Cut Your Fable 5 Bill by 70% With One Proxy</title><link>https://hqman.me/blog/pxpipe-token-hack/</link><guid isPermaLink="true">https://hqman.me/blog/pxpipe-token-hack/</guid><description>A local Claude Code proxy that renders bulky text as images so Fable 5 bills pixel tokens instead of 92k-character tool dumps.</description><pubDate>Sat, 04 Jul 2026 00:00:00 GMT</pubDate></item><item><title>MinerU: PDF to Markdown with OCR, Fully Local</title><link>https://hqman.me/blog/mineru-ocr-local/</link><guid isPermaLink="true">https://hqman.me/blog/mineru-ocr-local/</guid><description>Convert PDFs to Markdown with LaTeX formulas and extracted images on your own machine. MinerU runs offline with OCR for 109 languages.</description><pubDate>Fri, 03 Jul 2026 00:00:00 GMT</pubDate></item><item><title>Fable 5 Is Back, Sonnet 5 Costs More Than You Think</title><link>https://hqman.me/blog/fable-5-sonnet-5-builder-update/</link><guid isPermaLink="true">https://hqman.me/blog/fable-5-sonnet-5-builder-update/</guid><description>Fable 5 reopens on July 1 and Sonnet 5 is the new default. Usage, pricing, and why the cheaper sticker can still cost more.</description><pubDate>Wed, 01 Jul 2026 00:00:00 GMT</pubDate></item><item><title>Your Agent Doesn&apos;t Need a Bigger Window. It Needs Context Governance.</title><link>https://hqman.me/blog/headroom-context-governance/</link><guid isPermaLink="true">https://hqman.me/blog/headroom-context-governance/</guid><description>Headroom sits between your agent and the model, classifying and compressing tool output before it fills the context window.</description><pubDate>Tue, 30 Jun 2026 00:00:00 GMT</pubDate></item><item><title>Turn Articles into Animated System Diagrams: Visual Flow GIF</title><link>https://hqman.me/blog/visual-flow-gif/</link><guid isPermaLink="true">https://hqman.me/blog/visual-flow-gif/</guid><description>A small skill that extracts system structure from an article, writes a JSON spec, and renders an animated GIF with Python and Pillow.</description><pubDate>Sun, 28 Jun 2026 00:00:00 GMT</pubDate></item><item><title>Agents can now deploy to Cloudflare without signing up</title><link>https://hqman.me/blog/cloudflare-temporary-accounts/</link><guid isPermaLink="true">https://hqman.me/blog/cloudflare-temporary-accounts/</guid><description>Wrangler --temporary gives an agent a 60-minute Cloudflare account, a live workers.dev URL, and a claim link. No OAuth, no MFA, no human click.</description><pubDate>Fri, 26 Jun 2026 00:00:00 GMT</pubDate></item><item><title>GPT-5.5 Silently Throttles Reasoning for Third-Party Clients</title><link>https://hqman.me/blog/gpt55-reasoning-throttle/</link><guid isPermaLink="true">https://hqman.me/blog/gpt55-reasoning-throttle/</guid><description>Third-party GPT-5.5 clients can cap reasoning at 516 tokens. You pay full price and may get a weaker answer.</description><pubDate>Wed, 24 Jun 2026 00:00:00 GMT</pubDate></item><item><title>Codex Is Burning Through Your SSD. Here&apos;s How I Stopped It</title><link>https://hqman.me/blog/codex-ssd-disk-wear-fix/</link><guid isPermaLink="true">https://hqman.me/blog/codex-ssd-disk-wear-fix/</guid><description>A practical guide to checking Codex&apos;s SQLite log writes, blocking the unwanted inserts, and protecting your SSD without deleting conversation history.</description><pubDate>Tue, 23 Jun 2026 00:00:00 GMT</pubDate></item><item><title>GLM 5.2 vs Opus 4.8: Frontend Dashboard Test</title><link>https://hqman.me/blog/glm-52-vs-opus-48-frontend-test/</link><guid isPermaLink="true">https://hqman.me/blog/glm-52-vs-opus-48-frontend-test/</guid><description>A head-to-head frontend test of GLM 5.2 against Opus 4.8, plus Design Arena rankings, pricing, and the caveats that still matter.</description><pubDate>Sun, 21 Jun 2026 00:00:00 GMT</pubDate></item><item><title>Codex Can Now Copy Your Actions</title><link>https://hqman.me/blog/codex-record-replay/</link><guid isPermaLink="true">https://hqman.me/blog/codex-record-replay/</guid><description>Codex Record and Replay watches a Mac workflow once, turns it into an editable Skill, then reruns it with new inputs. macOS only, not in the EU.</description><pubDate>Sat, 20 Jun 2026 00:00:00 GMT</pubDate></item><item><title>Doubao&apos;s Full Agent Skill Architecture Leaked: 25 Skills, 283 Files</title><link>https://hqman.me/blog/doubao-agent-skills/</link><guid isPermaLink="true">https://hqman.me/blog/doubao-agent-skills/</guid><description>A leaked Doubao skill pack shows 25 modular skills and 283 files organized around real work objects, not one giant agent.</description><pubDate>Thu, 18 Jun 2026 00:00:00 GMT</pubDate></item><item><title>40 agent loops you can copy into Claude Code or Cursor</title><link>https://hqman.me/blog/agent-loop-templates/</link><guid isPermaLink="true">https://hqman.me/blog/agent-loop-templates/</guid><description>loops.elorm.xyz packages 40 agent loop templates by category so you can copy a kickoff prompt into Claude Code, Cursor, Codex, or Gemini CLI.</description><pubDate>Tue, 16 Jun 2026 00:00:00 GMT</pubDate></item><item><title>Fable 5 Is Gone. The System Prompt Isn&apos;t.</title><link>https://hqman.me/blog/fable-5-system-prompt/</link><guid isPermaLink="true">https://hqman.me/blog/fable-5-system-prompt/</guid><description>Claude Fable 5 lasted about 72 hours. The leaked system prompt is still public, and you can load it onto another model.</description><pubDate>Sun, 14 Jun 2026 00:00:00 GMT</pubDate></item><item><title>How to make AI-assisted writing sound less like slop</title><link>https://hqman.me/blog/humanizer-skill/</link><guid isPermaLink="true">https://hqman.me/blog/humanizer-skill/</guid><description>Humanizer is an open-source agent skill that catches 33 AI writing patterns. Pair it with your own voice samples and ship cleaner drafts.</description><pubDate>Sat, 13 Jun 2026 00:00:00 GMT</pubDate></item><item><title>Fable 5 One-Shotted This 3D Globe Visualization</title><link>https://hqman.me/blog/fable-5-one-shot-globe/</link><guid isPermaLink="true">https://hqman.me/blog/fable-5-one-shot-globe/</guid><description>One screenshot and one prompt. Fable 5 rebuilt a production-looking Three.js globe dashboard in a single pass.</description><pubDate>Thu, 11 Jun 2026 00:00:00 GMT</pubDate></item><item><title>Claude Fable 5 Is the Best Model You Can Use Right Now. Save It for the Hard Stuff.</title><link>https://hqman.me/blog/claude-fable-5/</link><guid isPermaLink="true">https://hqman.me/blog/claude-fable-5/</guid><description>Claude Fable 5 leads on hard coding benchmarks, but the cost is high. Send the hard work to it and keep everyday tasks on cheaper models.</description><pubDate>Wed, 10 Jun 2026 00:00:00 GMT</pubDate></item><item><title>Loop Engineering: A 14-Step Roadmap from Prompter to Loop Designer</title><link>https://hqman.me/blog/loop-engineering-14-step-roadmap/</link><guid isPermaLink="true">https://hqman.me/blog/loop-engineering-14-step-roadmap/</guid><description>A practical roadmap for deciding whether to build an agent loop, then adding automation, state, verification, tools, and security controls in the right order.</description><pubDate>Tue, 09 Jun 2026 00:00:00 GMT</pubDate></item><item><title>Loop Engineering: The Next Layer After Prompt Engineering</title><link>https://hqman.me/blog/loop-engineering-next-layer/</link><guid isPermaLink="true">https://hqman.me/blog/loop-engineering-next-layer/</guid><description>Peter Steinberger and Boris Cherny both say the job is designing loops, not writing prompts. Feedback, stop conditions, and token cost decide if it works.</description><pubDate>Tue, 09 Jun 2026 00:00:00 GMT</pubDate></item><item><title>A Practical Guide to Using /goal in Codex</title><link>https://hqman.me/blog/codex-goal-completion-contract/</link><guid isPermaLink="true">https://hqman.me/blog/codex-goal-completion-contract/</guid><description>Use Codex goals as completion contracts with explicit outcomes, verification, constraints, iteration rules, and safe stop conditions.</description><pubDate>Fri, 05 Jun 2026 00:00:00 GMT</pubDate></item><item><title>Codex Is No Longer Just a Coding Tool</title><link>https://hqman.me/blog/codex-industry-plugins/</link><guid isPermaLink="true">https://hqman.me/blog/codex-industry-plugins/</guid><description>OpenAI repositioned Codex for knowledge work with role-specific plugins, hosted Sites, and Annotations. Non-developers are growing fastest.</description><pubDate>Wed, 03 Jun 2026 00:00:00 GMT</pubDate></item><item><title>MiniMax M3: Opus-Level Coding, DeepSeek Pricing</title><link>https://hqman.me/blog/minimax-m3/</link><guid isPermaLink="true">https://hqman.me/blog/minimax-m3/</guid><description>MiniMax M3 launched with 1M context, near-Opus coding scores, DeepSeek-like API pricing, and two ways to try it immediately.</description><pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate></item><item><title>Used Codex to figure out what was eating my Mac&apos;s storage</title><link>https://hqman.me/blog/codex-storage-cleanup/</link><guid isPermaLink="true">https://hqman.me/blog/codex-storage-cleanup/</guid><description>A read-only Codex prompt that turned macOS System Data into a prioritized storage map, with 180 GB reclaimable on a 1TB Mac.</description><pubDate>Tue, 26 May 2026 00:00:00 GMT</pubDate></item><item><title>Let Codex Distill Your Own Workflow</title><link>https://hqman.me/blog/codex-self-audit-workflow-community/</link><guid isPermaLink="true">https://hqman.me/blog/codex-self-audit-workflow-community/</guid><description>How Codex can review your recent work, find repeated manual workflows, and turn the best ones into reusable skills, agents, or automations.</description><pubDate>Mon, 25 May 2026 00:00:00 GMT</pubDate></item><item><title>Recreate the Viral 3D Medical Teaching Model in 4 Steps</title><link>https://hqman.me/blog/3d-medical-teaching-model-4-steps/</link><guid isPermaLink="true">https://hqman.me/blog/3d-medical-teaching-model-4-steps/</guid><description>Rebuild a classroom-ready 3D anatomy viewer with AI-generated images, 3D reconstruction, GLB compression, and a coding agent.</description><pubDate>Fri, 22 May 2026 00:00:00 GMT</pubDate></item><item><title>How to Build a Self-Improving Company with AI</title><link>https://hqman.me/blog/self-improving-company-ai/</link><guid isPermaLink="true">https://hqman.me/blog/self-improving-company-ai/</guid><description>YC partner Tom Blomfield&apos;s framing: make company knowledge machine-readable, then run loops that sense, decide, act, and improve overnight.</description><pubDate>Thu, 21 May 2026 00:00:00 GMT</pubDate></item></channel></rss>