Time to hit the beach. Opinions are my AIs.
State of Inference
Joined March 2011
- Tweets18.1K
- Following303
- Followers4.6K
- Likes10K
legend fr
one of the coolest products of all time
Yesterday was my last day at Anthropic.
I created Claude Code on desktop because I believe everyone can be a builder. Every feature request and bug report we got from this community was a gift that impacted millions of lives. I’m eternally grateful for those gifts, the product wouldn’t be what it is without you.
Anthropic is doing some of the most important work of our time. Contributing to that mission will forever be a career highlight for me. I never imagined I’d get to contribute to something so special in my lifetime.
Walking away from my best friends, a product I poured my heart into, and the mission was the hardest thing I’ve done in my career. But now I have some free time for GTA VI.
I'm already itching to get back to addressing all your feedback. I'll share what's next soon!
99% of people out there don't know you can run opus 5.5 from sunup to sundown all day err day on $200 / month, wild
no one will be fired or held accountable for any of this shit
"we're sorry we stored your sensitive data in a server from the stone age, anyways, your next lab appointment has been scheduled!"
"There are 21.1 billion active connected IoT (Internet of Things) devices globally as of late 2025." - @AnalyticsIoT
"Pointing out the architectural flaws, Poretsky highlighted that because many devices are built with highly specific, low-power tasks in mind, they "lack powerful processing power" which ultimately "leaves them vulnerable" - Scott Poretsky
yup we are giga-fucked, amigos
🔴Drop this in Gemini Argon now:
"Audit my Gmail and connected Google Workspace accounts against the recent "Black October" vulnerability wave. Cross-reference every service I currently use or have touched in the past and give me a complete remediation checklist: where I need to rotate API keys, revoke OAuth tokens, patch versions, or lock down MFA. Do not miss anything.
Threat Vector Reference
LiteLLM: Poisoned with a credential stealer (3.4M daily downloads). Flag any local or cloud projects, dev environments, or agent pipelines where I have pulled or tested this package.
Node.js (v22, v24, v26): High-severity RCE flaw. Identify any connected hosting, serverless functions, or CI/CD pipelines running these runtimes.
Railway: Platform compromise. Check for any past or active Railway deployments and flag every linked OpenAI, Anthropic (Claude), Stripe, or database key that needs immediate rotation.
GitLab: Critical 10/10 severity vulnerability across all versions since 2014. Check for any connected repos, personal access tokens, or webhooks.
PHP (9.2/10) & WordPress: Core PHP vulnerability and compromised plugin signing keys planting backdoors. Flag any legacy domains, client sites, or hosting accounts tied to my email.
Java: Runtime compromise. Flag any dependent services or legacy tooling in my stack.
Signal: Client bug leaving mic and camera hot after hanging up. Note version checks needed across devices.
Required Output
Immediate Key Rotations: Exact services (OpenAI, Claude, Stripe, GCP/Firebase, Cloudflare, etc.) found in my emails or docs that intersect with these breaches.
Account & OAuth Hardening: Every legacy or active account found in Gmail that needs MFA enabled, sessions killed, or Google OAuth access revoked."
Runtime & Dependency Patches: Specific repos, local environments, or cloud workers I need to update or scrub right now."
they added 600 new tests since yesterday
guys what doin
The biggest lie in the personal AI agent space right now is that they're setting you free
Freedom to ditch your accountant, tax prep, pizza ordering app (lol)
In reality, they're locking you into a deeper and more complex walled garden with ever increasing layers of financial and technical opacity
We don't want a new apple vs android situation, we want the walls to come down at the same time the utility and value increases
my agent: 6039 tests passing, and
me: WOAHA WOAH, wtf are you guys actually doing in there
my agent: good catch, 3402 tests are no longer relevant, 1099 are tests that test the tests, probably only 328 are legitimate, should I delete
me: no fuck it, just -imkinda tired today, y'all do your thing
utter nonsense for anyone that isn't part of the AI-pilled tech minority on X imo
soon tho maybe, soon
