After 16 years, a developer quit Gmail over its AI: 'it thinks I can't write my own email'

Wed, 03 Jun 2026 16:28:32 GMT

What finally made a 16-year Gmail user leave

Gmail's AI compose prompts overlaid on the inbox

The original post isn't about price or reliability — it's about tone. After 16 years on Gmail, the author moved to Fastmail on a custom domain because the product started behaving as if he were incapable of handling his own email: message summaries generated without being asked, reply drafts inserted automatically, and a persistent prompt to hit / for Help me write and Tab to 'improve' what he'd already written.

His thesis lands in one line: "The message you're sending is that you think I'm not capable of reading and writing my own emails." He calls it software that feels like it's actively trying to be disrespectful.

The features you can't quietly turn off

The sharper complaint underneath the rant is about consent. Several of the AI behaviors can't be disabled in isolation — switching them off means giving up long-standing, genuinely useful functionality like thread categorization in the same toggle. So the choice isn't 'AI on/off', it's 'AI plus the features you actually liked, or neither'.

The thread: 'if it were useful, they wouldn't have to force it'

The 733-comment thread kept circling one argument. As one reader put it, if these features were as desirable as the marketing says, they wouldn't need to be promoted this aggressively — the adoption would be organic.

"I can appreciate LLMs for some use cases, but writing emails for the user is the one that really baffles me. The amount of native speakers reaching for it is wild." — a commenter unconvinced by AI compose

A few people pushed back on the absolutism: a one-click "that time works for me" reply is fine; what grates is the editor second-guessing prose you already wrote.

The kill switch, and the exits

The most-upvoted practical reply isn't a rant at all — it's a fix. A commenter pointed to the off switch most people miss: Gmail Settings → General → disable the global smart features option, which strips out a large chunk of the AI at once.

For those leaving anyway, a Fastmail user in the thread made the unpaid pitch: app passwords, hide-my-email, iOS integration — feature parity with Gmail, minus the nagging. Whether 'turn off one setting' or 'change providers' wins, the signal for Google is the same: the AI nudges are pushing at least some of its most technical users toward the door.

Lumafield CT-scanned a BYD, and an owner in the thread caught the teardown getting a detail wrong

Wed, 03 Jun 2026 16:28:31 GMT

What Lumafield put under the CT scanner

CT scan render of BYD components

Lumafield's Scan of the Month put parts from a BYD through an industrial CT scanner — the same non-destructive imaging used for aerospace inspection — to look inside the battery pack, the switchgear, and the key fob without cutting anything open. The pitch: see, layer by layer, how a car from the world's largest EV maker is actually assembled.

'From the lithium mine to the port': the vertical-integration claim

CT cross-section of the BYD battery module

The writeup's headline argument is about integration, and one commenter pulled the key line straight out of it:

"The last company to vertically integrate a car from raw material to finished product at this scale was Ford. Today BYD's system runs all the way from the lithium mine to the port."

The same commenter added the caveat the article glosses over: both BYD and Tesla claim to make roughly 75% of their components in-house, so 'lithium mine to port' is a vivid framing more than a unique feat.

The thread: an owner corrects the key-fob detail

CT scan of the BYD key fob

The scan's caption described the key fob as having "a mechanical backup key, a flat metal blade in a hinged housing." A BYD owner corrected it: the key isn't hinged — the whole mechanical blade pulls out once a small clip is unlatched. A small thing, but exactly the kind of error a CT image invites and an owner catches.

Corroboration on build quality came from a reader whose friend, an HV-certified technician, is mid-autopsy on a BYD Shark: he came away impressed by how heavy-duty the construction is.

What the scans don't settle: scale, robots, and the data question

A Berlin-based backend developer noted the part the imagery can't show: BYD now ships cars by the millions and leans hard on factory robotics, so any systemic quality problem would surface at that volume — the opposite of the 'cheap labor' assumption.

And a skeptic in the thread flagged what the teardown leaves out entirely: the data these cars collect and the 'social credit' context they ship from — a reminder that an impressive CT scan answers the engineering question, not the trust one.

Beyond the headline — Hacker News, unpacked