Twitxy Saved

Sean Gearin

Verified account This account is verified. Learn more

@seangearin

~1,000 commits a month using Claude, ChatGPT, Grok, and Cursor. I tell you what works vs what's hype - with receipts. Building memory that follows you w/ Virgil
seangearin.co Joined July 2026
411 Following 55 Followers 162 Posts

Latest posts 5

Sean Gearin Verified account @seangearin
Every @claudeai @ClaudeDevs customer should be extremely upset by this. The business and support teams have now continuously misled customers on plan details. Something needs to be done here.

@thsottiaux @OpenAI confirmed that my 20x there is an actual 20x (as it should be).

I like Claude and worked heavily through their platform in self funding a business. I’ve been a (beyond loyal customer). There needs to be more accountability to your customers - support is almost entirely non-existent/useless. You need to bring some humans in. I have the businesss background - tell me where to send the resume. Change needs to start now @AnthropicAI
SataEric Verified account @SataEricUX Aug 30
The whole “20x” in Claude’s pricing is so misleading.

The $100 plan says 5x Pro limits, while the $200 plan says 20x Pro limits. That makes it sound like you’re getting 4x the usage.

But the 20x applies only to the 5-hour usage window. The weekly limit is basically just 2x the $100 plan.

I spent way too much time digging through the internet and Reddit to figure this out. They could’ve just stated the limits directly instead of marketing it as “20x.”

That alone would’ve saved me hours of confusion.
View quoted post on X
Sean Gearin Verified account @seangearin
My AI's memory had me filed under 11 different identities lol. Fixed it and it just landed upstream in gbrain.

Every time a conversation mentioned me slightly differently, the memory engine minted a new identity instead of checking the aliases it already had - 11 versions of me and 1000+ memories filed under a typo. I fixed it at save time so every new memory checks who it already knows first. Landed in v0.46.32 this morning, credited.
Sean Gearin Verified account @seangearin
X grades its public algorithm against a secret internal version. The public one has 3 possible verdicts - the secret one has 7.

Todays code drop included this new test harness that runs both on the same posts and logs every disagreement. The extra verdicts in the secret version are things like Downrank and Tombstone - ways a post gets buried that the published code can't even express. Four of the seven have nowhere to land in the open version. Hot off the press: github.com/xai-org/x-algo…
Sean Gearin Verified account @seangearin
A free unnamed model just caught a bug Grok 4.6 waved through.

Same PR, same prompt, 3 reviewers. Grok 4.6: ship. Codex: request changes. The mystery model (stealth/ox-alpha, free on OpenRouter till ~Aug 27): request changes, same missing flag - reddit_arctic.py:167, one-line fix.

6 of 6 on public repo tasks vs both. Nobody knows whose it is. You have my attention.