Publications

Work Orchestration

A practical blueprint for replacing status meetings with agents that share context, surface blockers, and route follow-ups on auto.

Autonomous Coding Agents

A CLI tool that lets coding agents ask questions with a timeout, using their best judgement if the human doesn't respond.

Work Orchestration

Anthropic's dynamic workflows move the plan out of the model's context and into executable code, making large-scale agent orchestration inspectable and reproducible.

AI Media Production

Reve 2.0 brings native 4K image generation, layout-aware editing, and low API pricing to creative workflows.

Autonomous Coding Agents

AI coding systems are evolving from inline completion into architecture intelligence tools that build and maintain operational models of entire codebases.

AI Media Production

A two-line root cause in the AI video pipeline: the LLM faithfully locked composition duration to the user's target, but ElevenLabs took as long as the words actually needed.

AI Media Production

Cloning an unfamiliar video framework and getting production-ready output with a few prompts, not because of the prompt, but because the skills encoded the structure.

AI Media Production

How the Trilogy AI CoE replaced a rigid template-fill video pipeline with an LLM-authored composition system that generates bespoke explainer videos tailored to each article.

Work Orchestration

A recursive yak-shave journey through local infrastructure optimization, proving that with AI at your fingertips, there's no excuse for tedious manual setup.

Agent Runtime Operations

The setup, the gotchas, and three Claude Code skills that do the remote OpenClaw install for you in 30 minutes.

AI Media Production

OpenAI's GPT Image 2 didn't just get sharper. It got smart — by abandoning the way image models used to work.

Autonomous Coding Agents

Front-end design standards and skills for getting good interface design from AI coding agents when you are not a designer.

Open Model Strategy

DeepSeek's V4 preview is a smaller news event than R1 was. It is also, quietly, a much bigger one.

Personal Agent Runtimes

Anthropic just took over the part of the agent stack everyone hates building. The price is a quieter kind of lock-in.

AI Governance & Auditability

Every inbound message gets five chances to be rejected. Here's why that's a feature.

Agent Federation Protocols

The protocol that started as a feature became something bigger when we stopped treating it like one.

Work Orchestration

The move from context orchestration to external execution in agentic systems.

Agent Federation Protocols

OGP's Project Layer creates shared workspaces across independent agents without breaking local boundaries.

Autonomous Coding Agents

GPT-5.5, Codex, and the developer layer Anthropic keeps underestimating.

AI Governance & Auditability

Why agent protocols need delegated authority, not just message transport.

Agent Security Boundaries

Credential brokering for agent security.

Personal Agent Runtimes

Agent Framework 1.0 is a big deal for enterprises. But the problem I actually have isn't an enterprise problem.

AI Media Production

Key demos from the launch livestream.

Agent Federation Protocols

Most organizations do not have a knowledge problem.

Open Model Strategy

Moonshot AI's Kimi K2.6 arrives at a convenient moment for agent builders: it is open, it is strong on coding benchmarks, and it treats multimodality as part of the main model rather than a side branch.

Agent Security Boundaries

Major Supply-Chain Impact Now Looks Probable.

Personal Agent Runtimes

Run npx agentize to have a turnkey agent a persistent memory, task ledger, and architectural rulebook into any repo in 60 seconds. Works instantly with Claude Code, Cursor, and OpenClaw.

Document & Knowledge Systems

Jay Khalife took Andrej Karpathy's LLM-maintained wiki idea and extended it into an operational system for customer strategy, simulation, and action.

Document & Knowledge Systems

Jay Khalife wasn't hired to build AI systems. He built one anyway, turning fragmented operational data into simulations, strategy, and a reusable pattern other teams could adopt fast.

Open Model Strategy

A same-day contrast between open local multimodal models and a closed frontier service.

Work Orchestration

Before writing a single line of code, spend two hours planning with Claude using dependency-aware task graphs, decision gates, and failure recovery cascades.

Personal Agent Runtimes

Codex, Claude Code, OpenClaw, and Hermes move AI from chat to action by giving agents the ability to reason and act inside bounded environments.

Agent Federation Protocols

Why agent federation doesn't need another platform, just a reliable way to say 'check this now' across boundaries.

Autonomous Coding Agents

Vercel open-sourced their reference background coding agent. Here is what to click if you are not an engineer, and what to copy if you are.

Autonomous Coding Agents

The Claude Code team built something that handles hallucination, context blowup, permission abuse, bash injection, and infinite retry loops. Here is what is actually in the source code.

Agent Federation Protocols

MCP is the tool layer, A2A is the agent interoperability layer, and OGP is the trust-and-coordination layer across gateways.

Autonomous Coding Agents

A searchable database of 1,166 exact code snippets and constants extracted from Claude Code's source, packaged as a zero-dependency CLI tool.

Work Orchestration

Introducing OpenSymphony, an implementation that uses Linear as the work source, OpenHands as the execution harness, and a Rust orchestrator to manage issue runs, workspaces, retries, and recovery.

Training & Adaptation

Google's Gemma 4 removes the cost barrier for custom enterprise models with Turbo Quant and per-layer embeddings, enabling fine-tuning on consumer hardware.

Agent Runtime Operations

A meta-debugging loop using Claude and OpenClaw to diagnose and mitigate regression crashes in OpenClaw 2026.4.5.

Agent Runtime Operations

A step-by-step guide to switching OpenClaw from Anthropic subscriptions to cheaper alternatives like Kimi K2.5 and OpenAI Codex.

Personal Agent Runtimes

A technical decomposition comparing OpenClaw's gateway-centric routing model with Hermes's learning-loop agent runtime.

Agent Federation Protocols

An honest post-mortem of 100+ OGP builds, covering public-key identity fixes, peer persistence bugs, and what actually works in agent federation.

Agent Reliability Evaluation

Strategies for reliable agentic workflows on a budget, including tool surface reduction, structured guidance, and hybrid model routing.

Agent Federation Protocols

What federated AI actually looks like when it stops being a demo: two VPs building a product together without ever messaging each other directly.

Work Orchestration

Using Beads with OpenClaw for dependency-aware agent orchestration that structurally prevents step-skipping.

Agent Runtime Operations

A deep dive into common OpenClaw memory and identity issues, with exact fixes for boot files, symlinks, overwrite protection, and behavior routing.

Agent Federation Protocols

Why AI agent skills can't solve cross-organizational collaboration, and why federated gateways are the missing protocol layer.

Agent Federation Protocols

A pragmatic comparison of Unix CLI tools versus MCP servers for AI tool integration, with a case for simplicity.

Multimodal Model Capabilities

How late-interaction retrieval and multimodal embeddings are reshaping the search stack beyond single-vector approaches.

Autonomous Coding Agents

Strategies for manipulating context windows, isolating token-heavy tasks, and lowering Cursor execution costs with alternative models.

Work Orchestration

Anysphere engineers demonstrate Cursor Automations, Custom Skills, IntelliJ integration, and end-to-end Jira pipelines.

Agent Federation Protocols

How Clawporate extends multi-tier gateway isolation into a production multi-tenant OpenClaw platform on AWS.

Agent Runtime Operations

Why credential bleed in single-gateway deployments demands tiered isolation, and how to split one gateway into five.

Work Orchestration

How an autonomous multi-agent system turns voice chats and brainstorms into live deployed applications.

Agent Runtime Operations

Nine Claude Code skills that standardize OpenClaw operations and eliminate configuration drift caused by ad-hoc changes.

AI Media Production

Open-weight music generation on consumer hardware, serverless OpenClaw on AWS Bedrock, and autonomous meeting-to-deployment pipelines.

Multimodal Model Capabilities

Alibaba's Qwen 3.5 packs 262K-token context and native multimodal reasoning into models as small as 0.8B parameters.

Agent Runtime Operations

A practical guide to running the GasTown multi-agent orchestration framework cost-effectively by swapping expensive Claude Opus workers for cheaper, capable models like GLM-5 and Kimi K2.5.

Agent Runtime Operations

Moving OpenClaw from a fragile local toy to a reliable production tool through hard-won lessons in deployment, security, and practical agent operations.

Agent Runtime Operations

Feb 26 Office Hours recap covering how to slash multi-agent token costs with GasTown, deploy Kimi Claw in 60 seconds, and why Intent Engineering is becoming the new standard.

Personal Agent Runtimes

The four architectural decisions that let Gastown sustain 20-30 autonomous agents working for days without human intervention: self-propelling work, ephemeral sessions, observable state, and AI patrol.

Personal Agent Runtimes

Feb 19 Office Hours recap diving into OpenClaw's situated agency architecture, building Chrome extensions with Claude in five minutes, and the shifting model landscape beyond Anthropic.

Enterprise Workflow Automation

A vision for a meeting copilot that uses one avatar seat and many specialist brains, powered by a summoning pattern that dynamically routes context to the right sub-agent.

Personal Agent Runtimes

Beyond the wrapper: the architectural decisions that make OpenClaw an actual execution environment rather than just another API wrapper with a tool loop.

AI Tutoring

Feb 12 Office Hours recap on uncapping Claude Code via AWS Bedrock to bypass rate limits, and rapidly assimilating new tech stacks through curated Learners Lens paths.

Agent Federation Protocols

Feb 5 Office Hours recap covering custom email agents with Brain Trust, local agent orchestration via Telegram, calendar triggers, and using MCP Hives externally in your IDE.

Open Model Strategy

A technical breakdown of Moonshot Kimi K2.5 as a multimodal coding heavyweight, with practical recipes for pinning it to Fireworks via OpenRouter across OpenCode, OpenHands, Claude Code, and Factory Droid.

Document & Knowledge Systems

How to transform a single-user prototype into a multi-tenant platform where Legal, Procurement, and HR teams view the same documents but extract different insights without seeing each other's data.

Agent Federation Protocols

A rebrand hijacking, 900+ exposed gateways, and the real cost of agentic convenience.

Enterprise Workflow Automation

The "One Week" Horizon and The Art of the $10 Million Dollar Day.

Autonomous Coding Agents

The Multi-Agent Orchestration Framework, The Visual Dojo, and The End of Terminal Hoarding.

Personal Agent Runtimes

Why Single-Agent AI is Legacy Tech: The Case for "Implicit Orchestration".

Autonomous Coding Agents

Methods for Optimizing File Tasks in Anthropic's Agentic Tool.

LLM Evaluation Methods

An experiment comparing three prompting strategies on a real database migration.

Enterprise Workflow Automation

Designing an AI-Driven Sales Risk Pipeline for the Enterprise.

Enterprise Workflow Automation

The "Singularity" CEO: Why the Era of the Specialist is Over.

Enterprise Workflow Automation

Office Hours Recap: How We Automated the research for the $100M sales pipeline, Center of Excellence new initiatives aimed at Business Units' assistance; plus the rise of markdown programming.

Enterprise Workflow Automation

Trilogy exclusive: combining TheAlgorithm and Braintrust to establish your X.com presence.

Agent Reliability Evaluation

A deep dive into Probabilistic Architecture, the 'Survey Room' method, and why ROI is the wrong metric for AI.

Document & Knowledge Systems

Building a contract intelligence platform that moves beyond basic text extraction to answer complex, hierarchy-aware questions.

Enterprise Workflow Automation

The 'Cyborg' approach to customer support: using AI to transform support agents into technical operators who solve problems.

Open Model Strategy

Granite, OLMo, Trinity, and Nemotron enter the ring as American labs mount a counteroffensive in open-weight AI.

Autonomous Coding Agents

Software architect turned ML researcher on why quantity beats quality, and why the best engineer might not know how to code.

Document & Knowledge Systems

Office Hours recap covering contract analysis progress, internal learning platforms, and the latest in social automation.

Enterprise Workflow Automation

The $100M handshake and the efficiency obsession: why the future is about turning one salesperson into ten.

Enterprise Workflow Automation

The human-in-the-loop: AI adoption, legacy systems, and critical decisions in enterprise customer relations.

Training & Adaptation

A post-mortem on why hybrid inference with RAG is the currently viable path for specialized domains after fine-tuning caused capability regression.

Document & Knowledge Systems

A technical post-mortem on detecting dormant RCE payloads, the data-backed decision to use Landing AI for legacy contracts, and how Braintrust is bringing asynchronous, collaborative agents to the team.

Enterprise Workflow Automation

A practical guide for moving from manual workflows to fully automated deploys using CI/CD, AI, and modern tools.

Training & Adaptation

Spec-based critique of Amazon Nova Forge’s replay buffer and RLVR claims, questioning whether the $100k premium is genuine moat or just operational convenience.

Document & Knowledge Systems

Highlights the economics of Amazon Nova Forge, a task force win on Legacy OCR with Landing AI, and why Windsurf outclasses Google’s Antigravity alongside the launch of CoE Assist.

Enterprise Workflow Automation

Practical fixes to turn ARBs from bureaucratic bottlenecks into streamlined, AI-assisted reviews—moving from calendar-driven gatekeeping to risk-based pipelines with automation.

Open Model Strategy

Covers DeepSeekMath-V2’s self-verifying math model, Prime Intellect’s INTELLECT-3 RL stack, and Ai2’s OLMo 3 full “model flow,” contrasting how each defines openness.

Document & Knowledge Systems

Covers Gemini 3 one-shot app builds, training your own GPT on free GPUs, and the realities of enterprise OCR.

Enterprise Workflow Automation

Shows how to classify changes by risk, automate CAB checks, and modernize change control with AI and pipelines.

Open Model Strategy

Covers Anthropic’s Claude Opus 4.5 launch and competitive positioning in coding and reasoning benchmarks.

Autonomous Coding Agents

Guide to designing CI/CD pipelines that ship fast without breakage, using AI-driven testing and opinionated stacks.

Enterprise Workflow Automation

Argues Bezos’ Project Prometheus signals the next enterprise wave: physical AI systems beyond chatbots.

Open Model Strategy

Survey of late Oct–Nov 2025 releases: SWE-1.5, Cursor Composer, MiniMax M2, Kimi K2 Thinking, Gemini 3, Grok 4.1, Antigravity IDE, GPT-5.1-Codex Max, and early signals like Penguin Alpha.

AI Media Production

Multi-stage pipeline for verifiable generative AI that enforces deterministic outputs in image generation workflows.

Training & Adaptation

Gemini 3 builds production apps in one shot and shows how to train your own GPT on free GPUs with minimal prompting.

AI Governance & Auditability

A moderation AI misclassified a human as synthetic, hiding 80–97% of replies—lessons on misdetection and platform trust.

Agent Security Boundaries

A record Azure DDoS attack as a warning on AI reliability, cloud fragility, and resilience planning beyond benchmarks.

Agent Security Boundaries

Dissects Anthropic’s GTG-1002 agentic system for cyber operations, highlighting architecture and security risks.

Open Model Strategy

Leonardo’s framework for rapid technical analysis plus universal model deployment at 90% lower cost.

AI Tutoring

LearnLens Chrome extension that turns YouTube into competitive intelligence for learning and GTM research.

Enterprise Workflow Automation

AWS Bedrock Agents, Cursor’s Composer, and why Kimi outperforms consultants on slides for production-grade delivery.

AI Governance & Auditability

Examines how AI systems increasingly learn from digital behavior patterns and the implications for human-in-the-loop design.

Open Model Strategy

Explores how next-gen chips and tight hardware–software integration unlock new performance ceilings for AI workloads.

Agent Reliability Evaluation

Why evaluating LLMs and agents is fundamentally broken—and how to make assessments that reflect real-world performance.

Enterprise Workflow Automation

How I built an AI engagement system for X by choosing robustness over perfection.

Open Model Strategy

How we cut generation time 43% by splitting our pipeline—three architecture decisions that made our Arabic education system work at scale.

Enterprise Workflow Automation

How AI can learn to speak in the language of engagement.

LLM Evaluation Methods

We built the perfect DSPy pipeline. It had validation, auto-correction, infinite loop detection. Yet, the smaller Falcon model still was unprepared to stand on its own.

AI Tutoring

How we're using DSPy to create an autonomous education engine that gets smarter with every question it generates

Open Model Strategy

A breakdown of Alibaba's Qwen3 suite, covering multimodal breakthroughs, agentic vision AI, and hyper-efficient model architectures.

AI Governance & Auditability

Why open-sourcing code without weights or data isn't true accountability, and how AI can turn transparency theater into real algorithmic audits.

AI Tutoring

How to read, question, and apply AI papers

Personal Agent Runtimes

First it was Dia, then came Comet. I downloaded Fellou.ai the other day, which bills itself as the first “agentic browser.” As I type this I’m also installing GenSpark’s new AI browser.

AI Media Production

Why Gemini 2.5 Flash Image marks a permanent shift in creative workflows

Autonomous Coding Agents

Genspark & Abacus Ship, Jules & Devin Slip

Work Orchestration

Kiro and the orchestration of multi-tool pipelines for human–AI teams

Document & Knowledge Systems

Willfully over-engineering a simple RAG bot to explore agentic workflows

AI Tutoring

Cost‑Aware, Format‑Strict, and Surprisingly Minimal

LLM Evaluation Methods

Quick Dive: An honest evaluation of where DSPy excels, what my implementation adds, and how you should (or shouldn't) use it

Training & Adaptation

A comparison of Agent Lightning, Handit.ai, and a Homegrown tool - AgentEvolve

Enterprise Workflow Automation

Always-on AI coaching that keeps every employee focused, sane, and one step ahead.

AI Media Production

Hands‑On with 2025’s AI Video Tools (and Why 8 Seconds Still Hurts)

Document & Knowledge Systems

From Concept to Production with AI-Powered Development

Multimodal Model Capabilities

The Simple Path from Text to Voice Avatar: Everyone can create a chatbot - I transformed any template-based chatbot into a visual, voice-enabled expert with complete control and scalability.

Training & Adaptation

Can RL loops continuously refine prompts, tools, and agentic pipelines?

Document & Knowledge Systems

How combining existing tools and best practices helped me tackle the challenge of discovering and validating educational resources at scale

Open Model Strategy

Meet the Three Musketeers of coding, reasoning, and instruction

LLM Evaluation Methods

From Satire to Scientific Measurement

Enterprise Workflow Automation

How to Build & Auto-Refine Algorithms Using Multi-Model LLM Loops

Agent Reliability Evaluation

How enterprises can extract valuable technical patterns from DeepAgent's sophisticated design while demanding empirical validation

LLM Evaluation Methods

Grok 4 vs. Kimi K2

Enterprise Workflow Automation

Enterprise Content Orchestration for Content Creation, Approval and Scheduling with n8n & Airtable

Document & Knowledge Systems

10-engine system learning from each run; LLM ‘orchestrator’; open source APIs for 2,000+ vetted resources; AI-driven build-while-learning approach—enhanced with Google’s GenAI Processors architecture

Autonomous Coding Agents

A Hybrid Approach to Software Quality

Document & Knowledge Systems

How to Tame Context Limits, Retrieve Structured Data, and Build Reasoning Agents for Enterprise-Scale Insights

AI Media Production

A modern workflow

Open Model Strategy

Opinion: From LangChain to Mem0, new benchmarks reveal million-token context windows plus a simple stack present a more compelling case than memory frameworks

Enterprise Workflow Automation

And a Four-Layer Framework for Scalable Enterprise Adoption

AI Tutoring

Trilogy’s 73% AI usage is industry-leading — but business value trails. Here’s how we’ll turn high adoption into measurable impact, with standards, proven wins, and a culture of continuous learning

Agent Federation Protocols

A hands-on breakdown of where MCP ends, where A2A begins, and why orchestration, not communication, is the real architectural battleground.

Personal Agent Runtimes

How a Fierce Debate Forged the Blueprint for the Next Generation of AI

Autonomous Coding Agents

A deep dive into its architecture, standout features, and where it still falls short

Enterprise Workflow Automation

Discover how AI-driven video analytics uncover, measure, and transform hidden workplace anti-patterns — translating rigorous research into actionable ideas for enterprise productivity and success

LLM Evaluation Methods

From Off-the-Shelf to Custom: A Benchmarking Study of Agentic Retrieval Pipelines

Agent Federation Protocols

Why AI Agents Can't Talk to Each Other (And How A2A Aims to Fix It)

AI Media Production

The era of single-model content creation is over. 20 minutes vs 120 minutes determines market leadership = 84% efficiency gain

Agent Federation Protocols

Model Context Protocol

LLM Evaluation Methods

Does agentic retrieval trump vanilla retrieval? What is the top performing combination of datastores and embeddings from a retrieval accuracy perspective

Autonomous Coding Agents

A Guide to Tools, Trust, and Transparency in AI Coding

AI Media Production

In todays market, 10 minutes vs 70 minutes determines who wins proposals. This 86% efficiency gain translates directly to competitive advantage worth $41,600 annual capacity per analyst.

Agent Reliability Evaluation

A comprehensive benchmark analysis of popular agentic frameworks including LangChain, LangGraph, CrewAI, and AutoGen, evaluating their performance in real-world scenarios and providing actionable insights for framework selection.

AI Media Production

From Theory to Practice with Automated Solutions

Agent Reliability Evaluation

Analysis of Framework Architectures, Capabilities, and Multi-Agent Dynamics

Agent Reliability Evaluation

Assessing Human Readiness and Synergies in Human-AI Evaluation

Agent Reliability Evaluation

Benchmarking Multi-Agent Coordination, Reliability, and Interoperability

Agent Federation Protocols

Enabling Seamless AI Agent Collaboration

AI Tutoring

The Future of Personalized and Self-Directed Learning

Educational Visual Generation

A Guide to AI-Powered Tools

Agent Reliability Evaluation

Beyond Manus AI

LLM Evaluation Methods

Creating Effective Datasets and Evaluation Criteria

Agent Federation Protocols

MCP Meets OVON in the Quest for True Interoperability

Open Model Strategy

Key Highlights and Developments

Document & Knowledge Systems

Leveraging a Knowledge Base for Continuous, Iterative Discovery

LLM Evaluation Methods

Proprietary and Open-Source Solutions

LLM Evaluation Methods

Overview, Comparison, and Recommendation

Document & Knowledge Systems

Bridging Structured Knowledge and Generative AI for Smarter Solutions