Can A.I. Already Do Your Job?

| Podcasts | February 18, 2026 | 42.5 Thousand views | 30:50

TL;DR

New 'agentic' AI coding tools like Claude Code and OpenAI's Codex now allow non-programmers to build complex software and websites through simple conversation, potentially displacing entry-level developers while the technology begins improving itself recursively.

💻 The Shift to Agentic Coding 3 insights

From manual coding to vibe coding

Andrej Karpathy coined 'vibe coding' to describe letting AI write software instead of typing code line-by-line, removing the requirement to know programming languages.

AI agents work autonomously

Modern agentic coding tools create project plans, select programming languages, and deploy teams of specialized sub-agents to handle research, building, and testing independently.

Accessibility expands beyond tech workers

Marketing, sales, and finance professionals now use these tools to automate workflows and build applications without engineering backgrounds.

🛠️ Tools Reshaping Development 3 insights

Claude Code's viral emergence

Anthropic engineer Boris Cherny created Claude Code as a side project by integrating the AI into terminal applications, leading to rapid adoption across the company and millions of users.

Complex builds in minutes

During the demonstration, Claude Code built a professional personal website with a playable Tecmo Bowl-style video game in under two minutes from a simple text prompt.

OpenAI's competing entry

OpenAI released Codex as its agentic coding system, joining Anthropic in driving the current acceleration of autonomous programming capabilities.

📉 Workforce Disruption 3 insights

Entry-level engineering jobs contracting

A Stanford study analyzing payroll records found employment for young software engineers has dropped approximately 20% from its 2022 peak.

Role transformation for senior developers

Experienced programmers now report spending their time supervising and orchestrating AI agents rather than writing code themselves.

Economic value through labor reduction

Companies that previously hired teams of five to ten developers may now need only one or two engineers managing AI tools to produce equivalent output.

🔄 Recursive Self-Improvement 3 insights

AI models building themselves

OpenAI's latest coding models were used to help build and improve their own training runs, creating a feedback loop where AI systems enhance subsequent versions.

Rapid capability gains

Hallucination rates have decreased while reasoning and complex problem-solving abilities have improved at a pace surprising even the technology's creators.

Verifiable output quality

Coding serves as an ideal test ground because success is binary—code either runs correctly or fails—allowing AI to debug itself more effectively than in ambiguous domains.

Bottom Line

Professionals should immediately learn to supervise and orchestrate AI coding agents, as the barrier to software creation has collapsed to natural language prompts, making traditional coding skills less essential while contracting entry-level opportunities.

More from New York Times Podcasts

View all
Patricia Cornwell on Her Dark Childhood and Best-Selling Novels
59:31
New York Times Podcasts New York Times Podcasts

Patricia Cornwell on Her Dark Childhood and Best-Selling Novels

Pulitzer Prize-winning author Daniel Krauss discusses his experimental World War I novel "Angel Down," written as a single circular sentence, and reflects on how horror literature provided emotional armor during his childhood and shaped his maximalist aesthetic.

1 day ago · 9 points
What the End of Spirit Airlines Means for the Future of Flying
31:56
New York Times Podcasts New York Times Podcasts

What the End of Spirit Airlines Means for the Future of Flying

Spirit Airlines' sudden shutdown marks the end of the ultra-low-cost carrier era that revolutionized U.S. air travel by unbundling fares and democratizing flight access, ultimately collapsing under the weight of legacy airline competition, rising labor costs, and a blocked merger that sealed its fate.

2 days ago · 9 points
Isabel Allende and Her Mother Told Each Other (Almost) Everything
45:39
New York Times Podcasts New York Times Podcasts

Isabel Allende and Her Mother Told Each Other (Almost) Everything

Author Isabel Allende reveals how writing 24,000 letters to her mother over three decades forged an unusually intimate bond and trained her literary voice, while reflecting on how her debut novel emerged from grief and exile to capture three generations of resilient women.

4 days ago · 9 points
Find Your Perfect Swimsuit
37:26
New York Times Podcasts New York Times Podcasts

Find Your Perfect Swimsuit

This episode breaks down how to choose swimsuits that actually fit changing bodies, explaining that brands using plus-size fit models create better proportions than those simply scaling up size 6 patterns, while specific fabric choices like textured or compressive materials address common concerns from postpartum bellies to long torsos.

4 days ago · 9 points