AI Agents & Vibe Coding

LangChain raised three hundred million dollars at a valuation of three billion, more than twice what it was worth a year ago. The company grew out of an eight-hundred-line side project by Harrison Chase.

The heart of the episode is how LangChain itself runs on agents. Chase estimates that coding agents cost roughly thirty percent of the engineering payroll, and explains how the company keeps that spending in check. A sales agent prepares email drafts for salespeople in Slack and learns from their edits. Another agent looks for new errors after each code deployment and proposes fixes on its own.

For your own startup, you will take away a ready design for an agent that finds customers and a way to keep model spending under control. We also look at how free open source code became a funnel for a paid service. The company does not disclose its revenue, and the two hundred fifty percent rise in conversion was measured without a control group.

Full transcript: https://arsentev.ai/podcast/langchain-3b-how-the-company-runs-on-its-own-agents

Host: Evgenii Arsentev, creator of Arsminal, founder of ArsentevAI, Dad4Kids and ReelsZavod.
Arsminal · ArsentevAI · Dad4Kids · ReelsZavod

Anthropic has almost broken even, while OpenAI spends about two dollars for every dollar it earns.

In November both companies' quarterly numbers came out: almost $17 billion in revenue for Anthropic, more than $10 billion for the first time for OpenAI. The headlines said Anthropic had overtaken its rival. But the two companies count revenue differently, and the gap is smaller than it looks.

Anthropic has filed for its IPO, and this is the most detailed look inside any company in the industry. Three quarters of its revenue comes from large customers, growth comes from the bottom up, from programmers using Claude Code, and advertising for the whole of 2025 cost $147 million. OpenAI has a huge funnel of free users, a small share of paying ones, a bet on advertising and computing commitments worth hundreds of billions.

If you build a product on someone else's models, this economy is what sets your prices and limits. Much of what we know about OpenAI's numbers comes from leaks and the press, and Anthropic does not disclose its customer acquisition cost; the episode covers these gaps too.

Full transcript: https://arsentev.ai/podcast/openai-anthropic-money-who-actually-makes-money-on-ai

Host: Evgenii Arsentev, creator of Arsminal, founder of ArsentevAI, Dad4Kids and ReelsZavod.
Arsminal · ArsentevAI · Dad4Kids · ReelsZavod

Eight people and AI agents built a real travel agency in under a year. Twenty-seven days after launch, OpenAI bought it.

Before Odessia, founder Francis Davidson spent ten years building Sonder: more than a thousand employees, hundreds of millions of dollars in leases, and a bankruptcy. He built the new company the opposite way. Agents handled not only the code but also marketing, customer support, legal and finance. They read meeting recordings and email and kept working at night while the people slept.

The episode covers the six rules the team worked by, what a startup can copy tomorrow, and which of the founder's numbers nobody has checked.

Sources: Francis Davidson, "OpenAI Acquires Odessia" (X and LinkedIn, Nov 10, 2026), Skift, TechCrunch, PhocusWire, TNW, BetaKit, Inc., Commercial Observer, LangChain video.

Full transcript: https://arsentev.ai/podcast/openai-buys-odessia-ai-agent-travel-agency

Host: Evgenii Arsentev, creator of Arsminal, founder of ArsentevAI, Dad4Kids and ReelsZavod.
Arsminal · ArsentevAI · Dad4Kids · ReelsZavod

You change the model, the prompt or how often the agent's context is cleared, rerun, and every test is green. That tells you less than it seems. In 36 runs of a coding agent under six context-clearing policies, 4,086 tests passed with zero failures while cost differed by 33.5% between the most and least expensive of the six condition averages. The agent wrote its own tests in every run, from 97 to 127 per run, so the runs had no shared checks to compare. This episode explains why a test suite can be blind to the very difference you are testing, why the tempting "longer sessions did less work" reading did not hold up, and the proposal that came out of a W3C Agent Conformance group discussion: every result set should say whether any check changed state (some, none or incomparable) and which ones did. Because a record that cannot say incomparable will say none, and a reader will hear equivalent. Paper on Qeios: https://doi.org/10.32388/0BV3Z8

Full transcript of this episode: arsentev.ai/podcast

Full transcript: https://arsentev.ai/podcast/ai-agent-tests-all-green

Host: Evgenii Arsentev, creator of Arsminal, founder of ArsentevAI, Dad4Kids and ReelsZavod.
Arsminal · ArsentevAI · Dad4Kids · ReelsZavod

An AI content engine is a system where AI writes the scripts, AI avatars appear on camera, a synthetic voice talks and software edits and publishes. This episode walks through a working one: built by one person with no technical background, 1,497 videos across 67 channels in 101 days, 52 cents per 1,000 views. What a human studio or influencers would cost for the same output, why organic reach grows with a delay, a forecast for 100 AI presenters, and four steps for a business. Technical report with every calculation: https://doi.org/10.5281/zenodo.22802228

Full transcript of this episode: arsentev.ai/podcast

Full transcript: https://arsentev.ai/podcast/ai-content-engine-cost

Host: Evgenii Arsentev, creator of Arsminal, founder of ArsentevAI, Dad4Kids and ReelsZavod.
Arsminal · ArsentevAI · Dad4Kids · ReelsZavod

Correction, 15 September 2026: the audio and the transcript below use the prices of report versions 1.0 and 1.1, which were wrong. At the correct prices, a run costs $2.17 when the context is cleared after every task, $1.62 at every third task and $1.78 when it is never cleared (not $2.68, $2.14 and $2.47). Clearing after every task costs 33.5% more than the best policy (not 25%). Never clearing costs 9.7% more (not 16%), and that difference is not statistically significant (p = 0.046; 0.14 after adjustment for multiple comparisons). A cache write costs 20 times a cache read, not twelve and a half. Corrected report: https://doi.org/10.5281/zenodo.22759217

Thirty-six runs, six session policies, and an answer that matches neither of the two popular rules

Full transcript of this episode: arsentev.ai/podcast

Full transcript: https://arsentev.ai/podcast/when-to-clear-agent-context

Host: Evgenii Arsentev, creator of Arsminal, founder of ArsentevAI, Dad4Kids and ReelsZavod.
Arsminal · ArsentevAI · Dad4Kids · ReelsZavod

A long episode for anyone who has never opened the black window full of text and is sure it isn't for them

Full transcript of this episode: arsentev.ai/podcast

Full transcript: https://arsentev.ai/podcast/ai-from-zero-vibe-coder

Host: Evgenii Arsentev, creator of Arsminal, founder of ArsentevAI, Dad4Kids and ReelsZavod.
Arsminal · ArsentevAI · Dad4Kids · ReelsZavod

How to cut your token spend in Claude Code, or any other AI, several times over. Why the model rereads the whole conversation on every reply, what the context actually is, and why a bloated context makes the model dumber. One command, one habit of writing to a file, and one rule: one task, one chat, closed.

Full transcript of this episode: arsentev.ai/podcast

Full transcript: https://arsentev.ai/podcast/claude-code-token-spend

Host: Evgenii Arsentev, creator of Arsminal, founder of ArsentevAI, Dad4Kids and ReelsZavod.
Arsminal · ArsentevAI · Dad4Kids · ReelsZavod