Click to read Build With AI, a Substack publication.Build with AI is a biweekly newsletter for engineers who want AI as a superpower, not a party trick. Agentic engineering playbooks, context & harness engineering, MCP deep dives, agent cost architecture — one usable mental model per issue. No hype, no filler.
No reviews yetBe the first to leave a review for Build With AI
Maker
📌
Hey Product Hunt! 👋
I'm Charu, one of the makers of Build with AI — from Packt, 8,000+ tech books deep, and written with subject-matter experts.
We started this newsletter because of a frustrating gap: AI gets you 70% of the way to working software almost instantly — and then you spend ten weeks rewriting what it built in ten minutes. Everyone's publishing hot takes about what AI can do. Almost nobody's writing about what it takes to actually ship with it.
So every week, we send engineers two usable mental model — no hype, no filler:
⚙️ The Agentic Engineering Playbook — a multi-part series on building production agents
🧠 Harness & context engineering — why your model isn't the bottleneck anymore
🔌 The MCP/A2A/ACP protocol stack — which protocol does what, and when to use each
🛠️ Claude Code deep-dives and agent cost architecture
This is for engineers already building — everything comes from practitioners working in production, not repackaged press releases.
It's free to subscribe. I'd love to know: what's the hardest part of shipping with AI for you right now? That question drives what we write next.
Thanks for checking us out! 🙏
Report
A searchable archive would be huge since you're publishing biweekly and the deep dives on MCP and cost architecture are the kind of thing I'll want to reference months later. Bonus points if you tag each issue by topic and stack so I can pull up just the agent harness ones when I need them.
Report
the focus on one usable mental model per issue is such a smart constraint, it forces the content to actually stick instead of turning into another firehose of half-baked takes
A searchable archive would be huge since you're publishing biweekly and the deep dives on MCP and cost architecture are the kind of thing I'll want to reference months later. Bonus points if you tag each issue by topic and stack so I can pull up just the agent harness ones when I need them.
the focus on one usable mental model per issue is such a smart constraint, it forces the content to actually stick instead of turning into another firehose of half-baked takes