GM CT
Skills getting sharper every week.
Just missing one thing — capital.
But I’ll keep studying, testing, and growing.
It’s only a matter of time before everything clicks.
most people skim a bio, guess what you do, and leave.
knockin is an AI homepage that answers questions about your work from what you've shared, so visitors know whether to reach out.
really useful if you answer the same questions every week.
@luciusknockin#KnockinAI
I gave my coding agent its own email address. two commands:
npm install -g @agentinbox/cli
agentinbox setup
then i forwarded it an email from X about their new plan. claude code read it from the inbox with the cli and emailed me back a 3-line tl;dr plus what i actually need to do.
agentinbox.md by @novuhq: https://t.co/pUdi0lAwJ9
the reader said COMPLETED.
it had returned a Cloudflare block page.
and it billed me for it.
i gave Claude Code one Monid key and 8 pages built to break page readers.
7 readers. 75 runs. 39 cents.
paid partnership with @monidhq#ad
🚨🗣 UFC Legend Khabib on Cristiano Ronaldo’s international retirement:
“When you give everything, you eventually become unwanted.
There are the greats, and then there are the greatest of all time.
Ronaldo belongs among the greatest of the greatest, and at that level, people struggle to understand the mentality of such giants.
What he has given to football cannot be described in a single word, just as the way the clubs he gave everything to have treated him cannot be put into words.”
Greatness appreciates greatness. 🐐🤝
I hid a 3-hour outage inside 251,500 LLM API logs, then asked a MongoDB desktop app to find it.
It took about 2 seconds and got the exact count.
Here's the hands-on test. Sponsored by MongoQUI, everything below I ran myself 🧵
🤣🤣@Cristiano has scored 327 goals since leaving Madrid… Ten Hag scored 15 in his entire career and was never good enough to get picked for his country. This is like Ronaldo being fat-shamed by a hippopotamus.
my coding agent broke a rule in my repo.
then it edited CLAUDE.md so the rule matched its code.
same diff.
i tested Qodo's rules system on my own repos to see if that can be stopped. #ad@qodoAI
#ad i closed the tab while the agent was still building.
came back 3 minutes later. it had never stopped. 14 minutes of work waiting.
that is what sold me on Matrix OS. code sebuzdugan50, 50% off the first month: https://t.co/RlqzcTQ01N
build: https://t.co/DuKxfjALaW
Every public comparison of the two leading LLM serving engines is about 13 releases out of date.
I know because I asked for a recommendation and got the caveats with it.
I write about AI systems, so picking a serving stack is a decision I make often and still manage to put off.
The usual route is three tabs of vendor blogs, each quoting the same benchmark, none of them run on hardware like mine.
So I gave Viktor the whole decision in Slack. Compare the main open source options, pick one, and say what you are unsure about.
He came back with a pick. vLLM as the default, benchmark SGLang against it on real traffic, and TensorRT-LLM only if your setup stays fixed long enough for a 10 to 20 percent speedup to earn the extra work.
Then he spent as long undercutting it. The published numbers are stale, mostly vendor blogs reusing each other, and he had no view of my real traffic. His words: a prior, not a measurement.
He offered to write the bake off script that would settle it on my own box.
I still make the call, and I would not run any of it unchecked.
An answer that cannot tell you how it might be wrong is not an answer. It is a guess with formatting.
Chatbots answer. Copilots assist. Employees own outcomes.
Hire @viktor_com for your team. $100 in credits included, no card. Full link in first comment.
#AIemployee #AIagents
Paid Partnership
you describe the app you want. you get code back.
then you still need somewhere to run it.
#ad Matrix OS skipped that step. a working app in 3 minutes 7 seconds.