Learn AI in Bits

Is your job safe from AI, or does the data already point the other way, and can you even trust the layoff headlines that say AI did it? Three separate research efforts point to which specific job titles are most exposed to AI right now, and the layoff headlines blaming AI get checked against whether that explanation actually holds up. Microsoft's research team analyzed two hundred thousand real conversations from Bing Copilot in 2024, mapping them against government job classifications to score how much each occupation's daily tasks overlap with what generative AI already does well. The World Economic Forum's Future of Jobs Report named accountants and auditors, alongside cashiers and administrative assistants, among the fastest-declining roles by 2030, while Microsoft's data pointed to customer service representatives and sales representatives as similarly exposed. On the other end, roofers and water treatment plant operators ranked among the least AI-exposed jobs studied, since physical, hands-on work still resists automation. Outplacement firm Challenger, Gray and Christmas tracked more than ten thousand three hundred U.S. layoffs through July 2025 explicitly attributed to AI, with tech sector cuts up thirty-six percent year over year.

But the layoff data cuts both ways. Some companies genuinely are redirecting payroll into AI infrastructure: Meta cut roughly eight thousand roles, about ten percent of its workforce, in May 2026, with Mark Zuckerberg tying the move directly to rising AI costs, part of a broader trend as Alphabet, Amazon, Meta, and Microsoft plan to spend roughly seven hundred fifty billion dollars on AI infrastructure in 2026. Other times, AI functions as a convenient label. Investor Marc Andreessen has argued many large companies are overstaffed from pandemic-era hiring by twenty five to seventy five percent, calling AI the easiest excuse for cuts that were coming regardless. Salesforce CEO Marc Benioff has said layoffs from several distinct causes get lumped together under the same AI headline. A Duke University and Federal Reserve survey of company financial officers found AI's real employment impact closer to zero point four percent of jobs nationwide, a much narrower slice than the headlines imply.

The full picture includes which jobs the underlying task data actually points to, plus a more skeptical framework for reading the next AI layoff headline.

Sources & References

Microsoft study on jobs most and least exposed to generative AI — https://www.geekwire.com/2025/microsoft-study-shows-jobs-most-and-least-impacted-by-generative-ai/

World Economic Forum, Future of Jobs Report 2025 — https://www.weforum.org/stories/2025/01/future-of-jobs-report-2025-the-fastest-growing-and-declining-jobs/

Challenger, Gray & Christmas 2025 layoff data — https://fortune.com/2025/08/07/summer-of-ai-layoffs-july-140-percent-spike-challenger-gray-christmas/

AI-washing of layoffs, Andreessen and Benioff commentary, Duke/Federal Reserve CFO survey — https://sfstandard.com/2026/04/02/ai-washing-layoffs/

Companies cutting jobs to fund AI investment — https://247wallst.com/investing/2026/05/26/companies-are-cutting-jobs-for-ai-but-not-the-billions-paid-to-investors/

Voice narration is AI-generated.

Does AI actually make coding faster? This Deep Dive works through three questions: how a model learns to write code in the first place, what a coding agent really does when you hand it a goal, and whether the productivity gains everyone assumes are real. The honest answer involves a randomized trial where developers were confidently wrong about their own experience.

This is a longer-format episode, running about twenty minutes instead of the usual five.

WHAT'S COVERED

Part one - How a model learns to write code. Why coding improved faster than almost any other AI capability, and why the reason has nothing to do with code being easy. Code has a property that essays and advice and analysis do not: it can be checked automatically, with no human judgment involved. We walk through both training stages, from predicting the next token across vast amounts of public code, to the feedback loop that trains a model on whether its code actually ran. Then the benchmark that shows the effect: SWE-bench, built from real bug reports in real open-source projects, where early models solved under five percent of tasks and leading models now score around eighty percent. Plus two honest limits on that number, including why passing every test is not the same as writing good code.

Part two - What a coding agent actually does. The four rungs of AI coding, from autocomplete through chat and editor integration to full agents like Claude Code and Codex. A worked example of an agent fixing a real bug: searching the codebase, reading the relevant files, making a change, running the test suite, seeing a failure, and correcting itself without being told. How this shifts a developer's job from writing toward specifying and reviewing, why instruction files inside repositories now matter, what vibe coding means, and the two genuine risks — agents that can run destructive commands, and review becoming the real bottleneck.

Part three - Does it actually work? Why the widely quoted statistics on AI-written code are close to meaningless, including the fact that published figures for the same moment range from thirty to seventy-five percent because nobody defines the term the same way. Then the most rigorous evidence available: a randomized controlled trial from METR in which sixteen experienced developers completed real tasks nineteen percent slower with AI, after predicting they would be twenty-four percent faster, and still believing afterward that they had been sped up. We give that study its proper caveats rather than treating it as a verdict, then look at where the gains are genuinely large, where they shrink, and what rising code churn and falling developer trust suggest.

IN THIS EPISODE

AI coding, coding agents, Claude Code, Codex, GitHub Copilot, Cursor, autocomplete, SWE-bench, reinforcement learning, verifiable rewards, vibe coding, METR study, developer productivity, randomized controlled trial, code churn, technical debt, code review, AI-generated code, software engineering.

Part of a series covering tokens, agents, MCP, retrieval-augmented generation, transformers, neural networks, parameters, hallucinations, and vector databases.

Voice narration is AI-generated.

Tools, connectors, and skills sound interchangeable, but they solve three different problems. This episode separates them: what a tool actually is and why the model never runs one itself, how MCP turned custom integrations into a shared standard adopted across the industry, and what a skill adds once a model can already reach your systems. 

In this episode: tools, tool use, function calling, connectors, skills, MCP, Model Context Protocol, agents, integrations, OpenAI, Anthropic, Linux Foundation, AGENTS.md, open standards. 

Part of a series covering tokens, agents, MCP, retrieval-augmented generation, transformers, and vector databases.Voice narration is AI-generated. 

What is a vector database, and how does a system find the right document when it shares none of your words? This episode explains how text becomes coordinates, why searching by meaning is really searching by distance, the speed trade that makes it practical at scale, and the trap of a search that always returns something. Voice narration is AI-generated.

What is a transformer, and why is it the T in GPT? This episode explains the 2017 design that reshaped AI: why reading a sentence all at once beat reading it word by word, how attention lets every token weigh which other tokens matter, and why that one idea made models both scalable and genuinely good at context. Voice narration is AI-generated.

What is a neural network, and why does this one idea sit underneath almost all modern AI? This episode keeps it simple: what a single unit actually does, how layers of them turn a handwritten digit into an answer, and why learning features from examples beat decades of humans writing rules by hand. Ends where transformers begin. Voice narration is AI-generated. 

What is a parameter, and what do those billion-parameter headlines actually count? This episode builds the smallest possible model out loud to show what a learned number is, then answers the questions that follow: where parameters live inside a model, how much they can hold, whether more of them makes a model better, and whether they ever change. Voice narration is AI-generated. ?

What exactly is an AI hallucination, and why do models state false things so confidently? This episode explains how hallucinations arise from next-token prediction, why recent research argues that grading models on accuracy actively rewards guessing over admitting uncertainty, what it has cost in real courtrooms, and why asking a model to check its own work doesn't work. Voice narration is AI-generated.

Can AI successfully pick stocks? This episode weighs what the evidence actually shows: studies where language models read news headlines and predicted the next day's moves, against the humbling record of real AI-run funds that have trailed the S&P 500 for years. Learn why a genuine signal still fails to beat the market, and where AI does help. Voice narration is AI-generated.

Will AI take your job? The honest answer starts with what the research measures — and most of it measures tasks, not job titles. This episode works through the major studies, separates exposure from displacement, and lands on where the evidence is currently strongest.

On exposure: a landmark study from OpenAI researchers estimated that around 80% of US workers could have at least 10% of their tasks affected by large language models, with about 19% seeing at least half their tasks affected. The International Monetary Fund puts roughly 40% of jobs worldwide as exposed, rising to about 60% in advanced economies. Both measure exposure, not predicted job losses.

The episode surfaces a striking tension inside Gartner's own research. Gartner forecasts that by 2029, agentic AI could autonomously resolve as much as 80% of common customer-service issues. Then in April 2026, Gartner surveyed service and support leaders and found 85% were expanding what their human agents do rather than shrinking it — the same firm, the same industry, with forecast and current practice pointing opposite ways.

Worker-level research points the same direction. In a study of more than 5,000 customer-support agents, generative AI raised productivity about 14% overall and 34% for novice and lower-skilled workers, spreading techniques that had belonged to the most experienced staff. On the macro side, the World Economic Forum projects roughly 170 million jobs created and 92 million displaced by 2030 for a net gain near 78 million — though driven by multiple macrotrends, not AI alone — while 39% of workers' current skills change or go stale.

The sharpest current evidence concerns younger workers. Stanford researchers, using payroll data covering millions of US workers through June 2026, found employment for 22-to-25-year-olds in highly AI-exposed occupations sitting about 19% below where it would be had it kept pace with less-exposed peers, with no comparable gap for more experienced workers. The effect operates mainly through reduced hiring rather than layoffs. The episode holds this against the productivity finding to surface an uncomfortable pairing: AI helps junior workers most, and junior workers are the ones being hired less.

Sources & References

GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models, OpenAI — https://openai.com/index/gpts-are-gpts/

Gen-AI: Artificial Intelligence and the Future of Work, IMF — https://www.imf.org/en/Publications/Staff-Discussion-Notes/Issues/2024/01/14/Gen-AI-Artificial-Intelligence-and-the-Future-of-Work-542379

Gartner Predicts Agentic AI Will Autonomously Resolve 80% of Common Customer Service Issues by 2029 — https://www.gartner.com/en/newsroom/press-releases/2025-03-05-gartner-predicts-agentic-ai-will-autonomously-resolve-80-percent-of-common-customer-service-issues-without-human-intervention-by-20290

Gartner Survey Finds 85% of Service and Support Leaders are Expanding Human Agent Responsibilities — https://www.gartner.com/en/newsroom/press-releases/2026-04-28-gartner-survey-finds-eighty-five-percent-of-service-and-support-leaders-are-expanding-human-agent-responsibilities-despite-expectations-of-mass-ai-layoffs

Generative AI at Work, Brynjolfsson, Li & Raymond, NBER — https://www.nber.org/system/files/working_papers/w31161/w31161.pdf

The Future of Jobs Report 2025, World Economic Forum — https://www.weforum.org/publications/the-future-of-jobs-report-2025/

Canaries in the Coal Mine? Stanford Digital Economy Lab — https://digitaleconomy.stanford.edu/publication/canaries-in-the-coal-mine-six-facts-about-the-recent-employment-effects-of-artificial-intelligence/

Challenger Report: Layoffs Fall, Hiring Picks Up, AI Leads for Fifth Straight Month — https://www.challengergray.com/blog/challenger-report-layoffs-fall-hiring-picks-up-ai-leads-for-fifth-straight-month/

Voice narration is AI-generated.