Treat the TMT sector as one plumbing system (telecom the pipes, media the water, technology the taps), and a question falls out of the analogy: when automation started running amok, which pillar sprang a leak first? They did not all cave at once. They fell in a clear order: telecom, then media, then tech. And the sequence is not an accident.

The rule underneath the order

A pillar caves to automation the moment its atomic unit becomes cheap to fake and directly convertible to money, on infrastructure that has no native way to prove origin or humanity. Where those three line up earliest, the flood arrives earliest. That single rule predicts the whole sequence.

Telecom caved first

The phone number is a single, universal, forgeable identifier on a network that authenticates nothing about a call’s origin. The unit, a call, was the cheapest thing to automate the earliest: automatic dialers were harassing households in the 1970s and 80s, before the web existed. It paid out immediately (scams and telemarketing), and the reckoning came first too: the Telephone Consumer Protection Act in 1991, then the Do-Not-Call Registry in 2003. VOIP later changed only the economics, making origination and number-spoofing nearly free.

Media caved second

Media’s unit is the impression, click, or view. It became cheap-and-monetizable only once advertising was automated into a real-time auction: programmatic buying, roughly 2009 onward. Then money sat one hop from a forged event: fake the impression, collect the ad dollar, on measurement run by the parties being paid. The reckoning landed 2014-2016: the ANA/WhiteOps *Bot Baseline* put bot fraud near $7.2 billion for 2016, and Methbot was faking roughly 137 million video impressions a day by late that year.

Why media before tech

This is the part worth defending, because tech is where automation *lives*. Two reasons media fell first:

• Monetization latency. A fake media impression pays now. A fake web request pays nothing until you convert it into an account, a credential, or a transaction, an extra step. Money reached media’s door first, so fraud did too.
• Welcome automation. Tech’s bots were invited: the web was built for crawlers, and indexing is the business model. Tech never experienced automation as abuse, and never grew antibodies, until the incentive flipped from indexing to impersonation.

Tech caved last, and is still caving

Tech’s unit is the request, session, or account. Because its automation was structurally native and mostly benign, the adversarial flip came late, and in two waves. First, identity became the valuable unit: credential stuffing and account takeover became the dominant web attack 2016-2020 (billions of credentials spilled in 2016; Akamai counted tens of billions of stuffing attempts a year by 2018). Now the second wave, autonomous agents that can act like people, is arriving, and it is the subject of a whole different reckoning.

The complication that proves the rule

Email spam is the bridge case. It runs on tech infrastructure but behaves like telecom: a cheap message to a universal, addressable identifier with no origin proof. Tellingly, CAN-SPAM and the Do-Not-Call Registry both passed in 2003. That does not break the sequence. It confirms the rule. Wherever the "cheap message to a universal identifier with no origin proof" pattern appears, it caves early. Tech’s *general* layer (sessions, accounts, agents) caved much later.

The order is a warning, not trivia. Each pillar’s fix previews the next one’s. Telecom already learned that you cannot filter your way out at the edge. You attest at the origin. Media and tech are still relearning it.

Shailin Dhar

The three pillars share one nervous system now, and automation is moving through all of it. Telecom just happened to have the cheapest unit, the most direct payout, and the weakest proof of origin, so it went first. Read the sequence as a map of where the next flood lands.