Bharat Sharma | Articles

Engineers became senior through work that AI now does for them: tracing bugs across systems, reading unfamiliar code, discovering why old decisions were made.

Meta's collapsed 'AI-native' restructuring is being read as either proof AI can't replace knowledge workers or proof Meta moved too fast.

As AI makes first-pass code cheaper, engineering value shifts toward framing the right problem, understanding system-wide consequences, verifying changes, and owning production outcomes.

AI reliably makes individual engineers more productive, and that part has become easy: the tools work and adoption happens largely on its own.

Agentic systems fail differently from chatbots because they can act: choose tools, sequence steps, and keep going without asking.

The cost of producing work has fallen sharply; the cost of establishing that the work is correct and safe to act on has not.

AI has quietly rewritten what organizations expect from engineers, but the leadership side of that bargain went unrevised.

Most AI adoption stalls because leaders treat it as a tooling upgrade rather than an operating-model shift.

Shift-left fails when it's treated as a mandate to test earlier rather than a redesign of where quality lives.

Résumés measure where someone has been; trajectory measures how fast they're moving and in what direction.

Developer friction is rarely the engineers' fault.