15:I[981,["658","static/chunks/658-a698693f56d0116a.js","183","static/chunks/183-6beeeb801c274aaa.js","474","static/chunks/app/%5Btype%5D/layout-631c4730b83c8223.js"],"Posts"] 16:T528,A little more detail: I'm specifically a project manager in software development, and I've definitely noticed how quickly AI related technologies are advancing.

It's not hard for me to see the writing on the wall for my own profession, especially for lower performers.

I've had good relationships with the engineers I've partnered with on projects. Generally I just try to stay out of their way, enable them to make their own decisions instead of being some kind of task master. I've also tried to prevent external groups from bothering them, remove blockers ahead of time or as quickly as possible if needed. The most tiring part of the job for me is the politics, but that is like... what I get paid to deal with and shield folks from in my view.

I think if AI related tooling is integrated effectively, then the need for compiling and sharing information on a project gets reduced significantly (if not eliminated outright).

Maybe fewer project managers will be needed (if any). That's probably a good thing honestly. There's a lot of project managers out there that are pretty terrible. (maybe even me sometimes!)

I'm doing some serious soul searching on whether to leave the profession entirely after 12 years, or whether to stick with it. Open to suggestions.17:T5ff,Hi HN,

I’ve been working with a lot of Retrieval-Augmented Generation pipelines recently and kept running into the same debugging problem.

When a RAG system produces bad answers, people usually blame the LLM. But in many cases the issue is somewhere in the pipeline itself.

Things like:

documents not chunked correctly embedding models mismatched retrieval not happening before generation context windows overflowing vector database configuration problems prompt injection exposure

These kinds of issues are surprisingly har

Falsehoods programmers believe about addresses (2013)

Prompt-caching – auto-injects Anthropic cache breakpoints (90% token savings)

I'm a project manager, to the engineers: how replaceable do you think my job is?

Parametricity, or Comptime Is Bonkers

Show HN: Bonsai – A tiny sandboxed expression engine for TypeScript

Evolving our real-time timeseries storage again: Built in Rust (2025)

Western AI models "fail spectacularly" in farms abroad

Elusive Cost Savings of Pre-fabricated homes

Why yes, you *can* agentically code wherever you SSH

Social media is an ant mill

If you can't trust AI with a credit card, just give it a prepaid card

The Road Runner Economy

Researchers have demonstrated magnetic particle imaging on humans

Iran Drone War Ode Model

My Claude Code Setup

Show HN: RAG Doctor – CLI tool to diagnose broken RAG pipelines

Web Rewind

Microsoft Hasn't Had a Coherent GUI Strategy Since Petzold

From Run Nation to Power Slap: New Sports Designed for Social Media

Stanag 5516M-STD-6016 TADIL-J J-Series schema reconstruction from public docs

Want to Win a New CanaKit Raspberry Pi 5 Starter Kit Pro?

Prompt engineering vs. context engineering: a practical guide for AI builders

Solder Ninja Pen: USB-powered soldering iron compatible with Weller RT tips

OSS Document Scanner

Show HN: Line, Language with Intuitive and Natural Expression

40 Years of Wireless Evolution Leads to a Smart, Sensing Network

Analysis → Implementation → Reflection – a practical technique for agentic AI

Phoenix Arizona is likely to see its earliest 100f(38c) day on record, in March

Automating the Ticket-to-PR Cycle for Power Platform Code Apps with Azure DevOps

Hacking the Job Interview

Show HN: Payment Hunter – AI-powered invoice reminders for freelancers