Weekly #50-2025: Anthropic's Bun Bet, the PM Drought & Seattle's AI Backlash

Source: Dev.to

🎙️ Click here to listen on Madhu Sudhan Subedi Tech Weekly →

Bun JavaScript runtime acquired by Anthropic

Anthropic has snapped up the Bun JavaScript runtime in a move that could reshape how AI writes and ships code, elevating Bun from a fast new kid on the block to “essential infrastructure” behind tools like Claude Code’s native installer. Bun remains open‑source and MIT‑licensed, offering a turbocharged bundle of runtime, package manager, bundler, test runner, and single‑file executable builds.

Under the hood, Bun uses JavaScriptCore instead of V8 and Zig for native code, giving it a distinctive technical edge and a rebel identity in the JS world. Creator Jarred Sumner, who helped raise about $26 million for Bun while it generated essentially zero revenue, now sees Anthropic as the long‑term home that can fund aggressive development as AI agents begin to write, test, and deploy more software.

A tension remains: Zig’s inventor enforces a strict “no AI/LLM” contributor policy even as an AI company leans heavily on the language. Developers are split between excitement that Anthropic might supercharge Bun into the default JS/TS runtime for AI‑era apps and anxiety that today’s free, community‑driven project could tilt toward paid features or strategic pivots decided in boardrooms rather than GitHub issues.

Read the full article

Where Do You Go If You Don’t Care About Growth?

What if climbing the ladder isn’t the goal? A junior developer questions the default path of “grow, get promoted, repeat,” arguing that many roles offer little real upside for individual contributors while reliably enriching management. Performance is often judged by opaque criteria, “training” can mean conformity, and even unglamorous roles demand 5+ years’ experience to fix legacy messes.

Is there room for software careers centered on stability and purpose over status? Small, non‑growth companies, personal projects, and open‑source work can be more meaningful tracks—maintainable work that helps users and doesn’t exist solely to maximize someone else’s return.

The takeaway for teams and hiring managers: not everyone optimizes for speed and scale. A rising cohort values fair pay, sane scope, and autonomy over bonuses and titles. The question isn’t how to accelerate growth; it’s whether the industry can make space for engineers who choose craft and contribution over corporate ambition.

Read the full article

AI Companies Are Hiring Fewer PMs: 34% Lower Share of Open Roles

Are product roles shrinking as LLMs reshape how teams build? Riso Group finds that AI companies list 34 % fewer Product Manager titles as a share of openings compared to other sectors. PM roles account for 2.3 % of AI postings vs. 3.2–3.8 % across DevTools, Consumer, Data, B2B, and Fintech, based on 8,803 deduplicated job titles scraped from 100 tech firms. The analysis excludes big tech but the gap is clear.

The implication is that AI‑first orgs may be shifting headcount toward model building, infrastructure, and data roles while expecting existing teams to absorb more product responsibilities. PM scope could trend more technical and system‑oriented, with fewer pure orchestration roles.

Read the full article

Everyone in Seattle Hates AI

Why are many Seattle engineers hostile toward AI right now? This essay points to a cultural shift inside big tech—especially Microsoft—where layoffs, forced adoption of underperforming Copilot tools, and “AI or nothing” org politics have bred resentment. Engineers were told projects failed because teams didn’t “embrace AI,” even while being required to use tools that often made work slower and worse. The result: top talent rebranded as “not AI,” stuck watching compensation stall and autonomy vanish.

The environment shapes beliefs: engineers start to think AI is useless and they aren’t qualified to work on it, creating a self‑limiting loop that hurts companies (innovation gets centralized and stifled), employees (careers stagnate), and local builders (anything labeled “AI” triggers scorn). Compared with San Francisco, where curiosity and permission to build still exist, belief in one’s agency becomes a competitive advantage.

The takeaway for leaders: stop using AI as a political shield, empower teams to ship and evaluate tools honestly, and rebuild trust before cynicism hardens into permanent resistance.

Read the full essay

Treat Test Code Like Production Code

Is test code getting a pass it shouldn’t? Mark Seemann argues that the same standards applied to production code—readability, DRY, refactoring, reviews—should apply to tests. Sloppy patterns like copy‑paste suites, commented‑out “zombie” code, arbitrary waits, and blocking async calls make maintenance harder and slow teams down. The point of good code isn’t the computer; it’s future humans who have to read and change it.

Practical guidance: use test‑specific best practices (e.g., xUnit Test Patterns), write descriptive tests without sacrificing DRY, and avoid duplication that leads to “shotgun surgery” when behavior changes. Narrow dispensations exist—test code can relax some security and platform rules when it never ships (e.g., hardcoded test credentials, skipping certain async context rules in .NET, allowing orphan instances in Haskell).

The takeaway for teams: treat test code as a first‑class citizen; it’s part of the system’s maintainability, not a temporary scaffold.

Read the full article