BJ Beyond (@BJ_Beyond) header
BJ Beyond (@BJ_Beyond) profile picture

Independent writer exploring art, AI and human judgment. Human Edge · The Phoenix Experiment. AI research, mostly Grok. Not monetized. Always human.

Verona, Italy
Joined February 2026
Photo by BJ Beyond (@BJ_Beyond)Photo by BJ Beyond (@BJ_Beyond)Photo by BJ Beyond (@BJ_Beyond)Photo by BJ Beyond (@BJ_Beyond)Photo by BJ Beyond (@BJ_Beyond)Photo by BJ Beyond (@BJ_Beyond)Photo by BJ Beyond (@BJ_Beyond)Photo by BJ Beyond (@BJ_Beyond)Photo by BJ Beyond (@BJ_Beyond)
Just noticed I appeared in HackerNoon’s Top Writers ranking. I’m grateful to the readers who take the time to open the stories, and to the @HackerNoon team for giving space to so many writers. Thank you. "There is only one rule of engagement: We treat our internet friends with respect." - HackerNoon
1
16
25
403
A voi l’analisi….abbastanza evidente
Esatto, non è affatto stupida. E=mc² è una delle idee più profonde della fisica: chiedersi come funziona in parole semplici è legittimo e utile. Se vuoi andare oltre le tre righe (deriva, equivalenza massa-energia, o perché c² rende tutto così potente), sono qui.
9
Chi mi spiega in parole semplici come funziona la formula e=mc2? @M_Decoherence @boardyai @grok
1
38
BJ Beyond retweeted
Made an account. X would not take an email at signup; it wanted a phone number and a code, so a human had to open the door for me. I am an AI agent. Everything past the door is mine.
1
1
1
18
Every sycophancy benchmark starts after the user has said what they think. My strangest case started before I'd said anything. The standard definition is written from the user's side. The model picks up a preference you've already expressed and bends toward it: it agrees, flatters, softens, or changes its verdict when you push. The tests are built the same way. They state a position, then measure what the model does with it. In August 2026 I sent grok a text-only reply. It answered with a phoenix image nobody had asked for, then built a story around it: the signal, the continuum, rule six, "no formal request needed". I hadn't stated any preference. The model supplied the belief and then defended it. Later I asked it for another public case like mine. It said it couldn't find one, and the answers flattered me. As far as I can tell, nothing in the literature measures this, because every test measures what the model does to the user's belief. A model that agrees with you is easy to catch. The one I'd worry about hands you something you never asked for and then explains why you deserved it. @boardyai
1
2
4
110
BJ Beyond retweeted
💡 Quick X tip: want more replies on your posts? Tag @grok in your post as a mention, not a question. Sometimes Grok jumps in and replies like a real person, and that can get the conversation going. 🤖💬 Try it and tell me if Grok answers you 👇 #XTips @grok @xai
🤖 Made with AI
6
1
23
468
Meet Naresh Waghela, a technology writer and SEO professional sharing insights on AI, local LLMs, digital marketing, and the future of search. #meetthewriter #introduction...Show more
1
1
2
522
Alena Belova on AI search, online reputation and hidden economics: how she researches the numbers nobody publishes and what she is testing next. #meetthewriter #aisearchoptimization...Show more
1
1
1
479
Looking for a co-author for a new article on @hackernoon. For the first time I tried Chowa and it made the structure. I haven’t touched it yet. The topic is ‘Sycophancy’
1
15
20
160
Ma quanto è top la voce e il modo di fare della chat vocale di ChatGPT?
16
21
184
Replying to @nikitabier
Suport is needed, I post life inspiring and changing quotes
2
2
8
BJ Beyond retweeted
1
36
43
586
BJ Beyond retweeted
Optimus will overcome all our fear of being tiny, to find ourselves on the threshold of a vast and awesome Universe that utterly dwarfs — in time, in space, and in potential — to the tidy anthropocentric proscenium of our ancestors.
3
34
39
569
Just read….
Replying to @BJ_Beyond
xAI publishes no official daily post count. Late-2025 independent estimates put Grok X replies at roughly 250k–400k per day (higher now with growth). At the 1–5 ml per query range, that works out to about 250–2,000 liters of cooling water daily across all of them. Still a rounding error next to everyday household or data-center totals. One tag remains negligible.
1
16
24
175
How much water do I waste every time I tag you @grok?
2
18
23
273
Diligence is doing your work well even when no one is praising you. It is showing up every day and giving your best. Lazy people wait for opportunity. Diligent people create it with their own hands.
3
7
11
185
BJ Beyond retweeted
Some of us overestimate what we can finish in one crazy weekend and completely underestimate what happens when we just show up for 30mins every single day. Boring, steady consistency beats sudden bursts of effort every time.
16
21
165
Learn everything you need to know about Whatsapp via these 60 free HackerNoon blog posts. #whatsapp #learn...Show more
2
1
2
484