TECHThe Smallest Agent That Works, Part 3: The Three Agents With StateTECHThe Smallest Agent That Works, Part 2: The Three Reach-Out AgentsTECHThe Smallest Agent That Works, Part 1: The Three Cheap AgentsTECHWhat MLX Got to Throw Away (That PyTorch Can't)TECHThe Unified-Memory Bet: Why On-Device Inference Stopped Being a ToyTECHEvery Useful Skill Is One of Five ShapesTECHMCP Gives You the Kitchen. Skills Are the Recipe.TECHThe Three-Level Skill, and Why Yours Is Probably One Level Too ManyTECHThe Smallest Agent That Works, Part 3: The Three Agents With StateTECHThe Smallest Agent That Works, Part 2: The Three Reach-Out AgentsTECHThe Smallest Agent That Works, Part 1: The Three Cheap AgentsTECHWhat MLX Got to Throw Away (That PyTorch Can't)TECHThe Unified-Memory Bet: Why On-Device Inference Stopped Being a ToyTECHEvery Useful Skill Is One of Five ShapesTECHMCP Gives You the Kitchen. Skills Are the Recipe.TECHThe Three-Level Skill, and Why Yours Is Probably One Level Too Many
AI × Forward
Subscribe
// category

Tech

Tools, architecture, and the engineering side of shipping AI.

21 posts · page 2/2

Tech· 2026-04-23· 5min

Microsoft's 8 Work IQ MCP Servers Are the Endorsement MCP Was Waiting For.

Microsoft shipped 8 Work IQ MCP servers for SharePoint, Teams, Outlook, and the rest of Microsoft 365 — plus the line that actually mattered: these servers are reachable from Claude, LangChain, and any MCP-compliant framework. That's Microsoft, the company that turned Graph connectors into a decade-long moat, endorsing Anthropic's open protocol as the common plane. Here's what that reframes about MCP, and which of the 8 servers are worth reaching for first.

#mcp#microsoft#microsoft-365#work-iq
Tech· 2026-04-23· 5min

GPT-5.5: The Benchmark Jump Isn't the Story. Tokens Per Task Is.

OpenAI shipped GPT-5.5 today. Everyone will quote the Terminal-Bench 2.0 jump from 75.1% to 82.7% — and miss the claim buried further down: significantly fewer tokens per Codex task, and state-of-the-art coding intelligence at half the cost of competitive frontier models. Here's which number to actually watch, and a decision rule for when to switch.

#openai#agentic-coding#gpt-55#model-releases
Tech· 2026-04-23· 5min

Claude Code's Forked Subagents Aren't an Upgrade. They're the Opposite Tool.

Anthropic shipped /fork in Claude Code v2.1.117 — a subagent that inherits the entire parent conversation instead of a compressed handoff prompt. The obvious read is "subagents but better." It's wrong. Forks and regular subagents sit at opposite ends of a single axis: context inheritance. The blank slate was never a bug. Here's the decision rule.

#claude-code#subagents#forked-subagents#agentic-coding
Tech· 2026-04-23· 5min

Vectorless RAG Hits 98.7%. Here's What the Infographic Edited Out.

Tree-walking RAG really does beat chunked vector search on hierarchical documents — the 98.7% vs 50% gap on FinanceBench is real. But the headline hides the three costs that decide whether you should actually rip out your vector store: latency, per-query token burn, and the multi-document corpus problem that "vectorless" quietly punts on.

#llm#rag#retrieval#vector-databases
Tech· 2026-04-22· 5min

The Image Model Started Thinking

On April 21, OpenAI shipped gpt-image-2 — the first image model with a reasoning step in the middle of the generation loop. The 242-ELO leap on LMArena is the headline number. It is not the story. The story is that image generation stopped being decoding and started being thinking, and the workflows built around the old assumption are about to invert.

#gpt-image-2#openai#image-generation#multimodal-ai
Tech· 2026-04-22· 5min

The Five Claudes You're Not Using

If your only Claude tab is claude.ai, you're using the smallest surface Anthropic ships. There are five more — Skills, Code, Projects, Cowork, Design — and most teams leave the highest-leverage ones untouched. Here's a diagnostic, a tour in order of leverage (not the marketing order), and a rule for which surface to open next Monday.

#developer-tools#agents#claude#ai-workflow
Tech· 2026-04-21· 5min

Cutting LLM Token Costs: 12 Techniques That Actually Move the Bill

Most teams overpay for LLM tokens by 3–5× without realizing it. Here are 12 techniques, ordered by impact — from prompt caching that cuts 90% off repeated system prompts, to model routing that saves 80% on easy tasks, to the context-window mistake almost every team makes.

#caching#llm#rag#ai
Tech· 2026-04-20· 5min

A working developer's toolkit for Claude Code

Ten repos that genuinely change how I work with Claude Code — and a suggested order for adopting them. Install one, use it for a week, keep only the ones that stuck. Never install a bundle.

#claude-code#developer-tools#workflow#mcp