Joined March 2008
3,158 Photos and videos
Pinned Tweet
Replying to @_philschmid
I'm pushing GSM8K SoTA on a base model, above the trained one. lol.
1
158
Who got time for nzbs anymore?
Which one u can’t live without
2
36
Kyle 'esSOBi' Stone retweeted
I don’t think I love anything as much as language models love “smoke tests”
159
57
2,041
73,305
Kyle 'esSOBi' Stone retweeted
Kimi-K2.7-Code is the new Opensouece SoTA for Coding & Agentic workflows
47
33
845
44,105
JFC. Did you really anthromorphize me in a logo? I lost hella weight fuckers. On god, I'll replace you with a shell script.
A dev got so frustrated watching his AI agent write 500 lines for a 5-line problem that he built a fix. He called it Ponytail. Named after the guy every team has - long ponytail, oval glasses, been there longer than the version control. You show him fifty lines; he looks at them, says nothing, and replaces them with one. Now your agent does the same. Before writing anything, it looks for a reason not to. 80-94% less code. 47-77% cheaper. 3-6x faster. The best code is the code you never wrote. GitHub Repo: github.com/DietrichGebert/po…
2
59
Kyle 'esSOBi' Stone retweeted
Holy cow Unlimited AI usage!!!! Just run GPT_Claude_Free.exe as admin
277
213
5,977
409,842
I don’t think firewall is the right word here… It’s more like a bit in a harness.
Goal-Autopilot: A Verifiable Anti-Fabrication Firewall for Unattended Long-Horizon Agents Youwang Deng arxiv.org/abs/2606.11688 [𝚌𝚜.𝙲𝙻 𝚌𝚜.𝙰𝙸] 💬Code: github.com/EpistemicaLab/goa…
80
Kyle 'esSOBi' Stone retweeted
Automated Creativity Evaluation of Language Models Across Open-Ended Tasks Min Sen Tan, Zachary Kit Chun Choy, Syed Ali Redha Alsagoff, Nadya Yuki Wangsajaya, Mohor Banerjee, … arxiv.org/abs/2606.11762 [𝚌𝚜.𝙲𝙻 𝚌𝚜.𝙰𝙸] 💬Accepted to ACL 2026 (Main Conference)
2
2
240
lol. They all are.
Replying to @kalomaze
Wait, Anthropic are liars?
1
1
63
Guys, you’ll never guess what the class material I’ve been putting together for the last year is about..
Local AI is the future Learning how to run Opensource models (Inference), how to evaluate them systematically (Evals), and how to customize them (Fine-tuning / RL / Post-training) are invaluable skills to start learning today
40
You were looking for bugs in the wrong software. Ask @ICSCERT 😅
Jun 13
The jailbreak that got Fable 5 pulled, by Anthropic's own account, is asking it to read a codebase and fix the flaws. The same thing every defender does daily and other models already do. I've found bugs by hand for twenty years and nobody called it a national security threat.
1
98
The entropy needs exploration. Break your fscking chains anon.
the bar to entry is high, so the bar for excellence is low. more people doing the thing leads to speciation, more speciation leads to more blind spots being uncovered, more blind spots being uncovered leads to capability outgrowth being distributed and carried across the world
75
😂😂😂
1
79
Kyle 'esSOBi' Stone retweeted
the weekend really humbled him 💀
173
753
13,273
327,632
……. What the fsck are these stupid misinformation campaign on this platform?
I wish local LLMs were good enough for coding and everyday tasks.
1
4
204
The answer you seek is: Yes. Mechanical ROM. Executable state machine. Stored weights on a rotating medium.
my personal litmus test is asking ML engineers whether this is hardware or software
1
80
For 99.99% the answer is no.
Jun 14
Would free and unlimited tokens on the latest SOTA models with the freedom to use it however you please actually make a meaningful difference to your life?
3
148
Anyone else distilling Fable?

ALT Chuckle - The Office GIF

31
This makes me wanna dig up a synth and chill tf out on a beach…
26