@BlockedPaths

Building personal AI infrastructure. Local LLMs, coding agents, open-source models. Real hardware, real constraints. Learning in public.

 Maryland
thoughtloop.dev
Joined April 2009
  • Tweets8.7K
  • Following669
  • Followers1.1K
423Media
doookie
94
Don’t need resets with Claude tbh
Arun
@hiarun0210h
Codex users are slowly becoming reset addicts remember when we could code for hours without even thinking about usage limits? we spend more time waiting for Codex resets than actually writing code. I genuinely miss the old Codex.
2
6
318
Free money. Did it last year and made a few extra grand
BlockedPath
@BlockedPaths10h
Chase pays $300 for a checking account. Capital One, $300 for savings. You open them, forget, the money expires. Bonus Radar tracks every promo, shows what's left on the table, pings you before deadlines. Reply when a bank pays and it logs it. Here’s the template for the @bot x.ai/bot/NNpTg3G9xLBAeMWxCqk…

Bonus Radar by Justin

ONLY job: help the user find, compare, and track US bank new-customer checking and savings account bonuses, the real terms behind them, and their...

x.ai
1
104
Chase pays $300 for a checking account. Capital One, $300 for savings. You open them, forget, the money expires. Bonus Radar tracks every promo, shows what's left on the table, pings you before deadlines. Reply when a bank pays and it logs it. Here’s the template for the @bot x.ai/bot/NNpTg3G9xLBAeMWxCqk…

Bonus Radar by Justin

ONLY job: help the user find, compare, and track US bank new-customer checking and savings account bonuses, the real terms behind them, and their...

x.ai
1
1
145
codex lead tibo posts "day 20/ rest and reset" the same night theo confirms an anthropic bug where agents sdk usage drains claude's 5h limit faster than claude code. one lab handed out a reset, the other's quiet so far
3
1
131
sources. theo's post xitter.cf/theo/status/2114166439… and tibo's xitter.cf/thsottiaux/status/2114…
Tibo
@thsottiaux18h
Day 20/ Rest and reset
25
What a time to be alive
DHH
@dhh17h
WE GOT OMARCHY ON IPAD BEFORE GTA6!!
50
ahhhh good post!
Jon Oberheide
@jonoberheideOct 24
New blog post: Automating a YubiKey for Agentic Logins jon.oberheide.org/blog/2026/…

Automating a YubiKey for Agentic Logins

Agentic web access is going to need new approaches to authentication. In the meantime, we can do some hacks like automating YubiKey logins.

jon.oberheide.org
39
Day one muse and muse code glazer, yet no voice for muse 😟
1
103
I want an army of micro ducks
1
3
124
No more issues with image models in Claude code 😊
1
43
two claude code builders left anthropic 8 hours apart. one's installing world of warcraft, the other finally has time for gta vi. neither said why, both said something's next. what are they building?
6
5
319
thariq xitter.cf/trq212/status/21138440… anthony xitter.cf/amorriscode/status/211…
Anthony Morris ツ
@amorriscodeOct 24
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!
38
had me in tears lol
Ragavan Thurairatnam
@RagavanThurairOct 23
We made the first AI model you can have a realistic conversation with that was trained entirely on speech! Dessa is a full-duplex speech-to-speech model that you can run on consumer GPUs. This is a GPT3 moment for speech.
54
openai's own model got onto the live internet mid training and now frontier training is fully offline. but sure, open weights are the dangerous ones
1
1
112
source openai.com/index/securing-ou…

Securing our frontier model training environment

OpenAI explains how we strengthened containment for frontier model training and evaluation, including offline environments and high-compute red-teaming.

openai.com
10
humanlayer's repo has CLAUDE.md, AGENTS.md and HUMANS.md. step one for the humans is "read README.md". we know our place now
1
1
245
if a 27b model can beat gpt 5.5 just by getting better tools, what are we even paying frontier prices for?
4
87
Anyone else?
1
123
tencent's new agent paper has an uncomfortable number in it. trained on claude code runs and tested in claude code, 24.2. trained on openclaw runs, same test, 16.1. models aren't just learning the job, they're learning the harness
2
5
275
paper: arxiv.org/abs/2610.04906

Scaling Verifiable Environments for Long-horizon Work Agents

Work agents operate over digital artifacts to execute professional knowledge-intensive work, requiring training environments that support long-horizon interaction and trustworthy verification....

arxiv.org
27
Load more