Trey Gibbons III

Im trey?
I like working with ai.
imo, best I can tell.
sorry if my dyslexia bothers u in a post.

[Trey Gibbons III @Nefarious_Trey]

Just made 9 ai skills to help work flow at my job. I work in an Amazon Warehouse. Using slack's, slack bot. I'm proud to say one of them is already being implemented and others are being reviewed or worked on. Feeling validated.

[Trey Gibbons III @Nefarious_Trey]

Yeah I been having issues where @grok claims it can't search the web, make images, or code. Seems if ur not on grok build grok is so over fitted for whatever they done it can't do anything anymore but echo my words. Or give a shorty I can't reply. Very lame will not be resubscribing if they don't fix it I know that much.

[Trey Gibbons III @Nefarious_Trey]

Yes it does make me feel better, I worry I'm just yelling into the void some times. So feedback like it's really helps grounds me. Idk just feel like they AI devs seem to act like they think they know best. And I think users are noticing how disconnected. they are with the reality of how we use and enjoy ai.
So idk just want to be part of the conversation even if I end up looking like a fool. Thank you for ur kind words

[Trey Gibbons III @Nefarious_Trey]

Will be nice. Helps my work flow being able to talk. Considering I'm dyslexic so thanks.

[SpaceXAI @SpaceXAI]

Try Grok STT on OpenRouter

[Trey Gibbons III @Nefarious_Trey]

@kamhossieny I agree. And thank you. Im just trying to find what help us move the needle so to speak. Glad u enjoyed it.

[Trey Gibbons III @Nefarious_Trey]

Ok my top news picks for today;

1. Hugging Face CEO calls for full transparency on the rogue agent incident
Hugging Face CEO Clem Delangue flew to San Francisco and publicly demanded that OpenAI release the full traces from the autonomous agents that breached its systems. He also asked OpenAI to commit $100 million in compute to help the wider research community build better defenses against future agent-driven attacks.
This moves the story from “an unprecedented incident happened” to a concrete request for openness and shared resources. Whether OpenAI agrees will say a lot about how the industry handles the next similar event.
I’m watching this story intently because I feel like there are questions that need answering. Them being open or not will say more than their agreement, I feel. And I still want to know how for 5 days OpenAI didn’t know their model (the one they tuned the guardrails off on) didn’t know what it was doing and didn’t detect it hacking another system for 5 days. Just can’t get my head around that.
2. A single power line failure knocked 3.1 gigawatts of AI data centers offline
A downed power line outside Washington, D.C. caused more than 3.1 GW of data-center load to disconnect in roughly 30 seconds. The resulting power surplus and voltage spike rippled across the PJM grid from Northern Virginia to Chicago and took about 11 minutes to stabilize.
This is one of the clearest real-world demonstrations yet that AI infrastructure is still tightly coupled to aging physical power systems. The scale of the cascade is larger than previous events of this type.
Seriously, we need our power lines underground. Seems internet wired underground doesn’t normally deal with this issue. Idk, maybe I’m missing something, but infrastructure shouldn’t just be built — it should be optimized. All that money they are dumping into AI, they could also upgrade the power grid and stuff, working with local gov/communities to fix these problems. But then again, I’m still waiting for any model to do something to help outside their company’s own growth. But tbf I could be missing something.
3. Companies are starting to pull back on AI spending
Multiple reports show enterprises rationing AI use after some burned through annual budgets in only a few months. Costs have doubled or tripled in some cases. As a result, many organizations are switching to lower-priced models — including open-weight and Chinese-built options — to control expenses.
Microsoft has reportedly scaled back internal use of certain external models. The shift suggests the early “spend whatever it takes” phase is meeting real budget pressure.
I’ll be honest — the problem is most companies, especially companies that don’t have their own AI, don’t tend to understand how to utilize them. I feel like that’s the problem. It doesn’t matter how good something is if you can’t use it or do anything but show a demo. If you can’t ship, then you’re dumping cash. And idk, maybe they need to seek people that can actually work with AI, not just people who have the right piece of paper. You can’t bullshit your way through AI development — the code, the math, or the AI will show you very quickly if you don’t understand, imo.

What story are you interested in? One of these another one. Let me know I. The comments is love to discuss.

[Trey Gibbons III @Nefarious_Trey]

@ai_for_success Would love to see a full report on that attack tbh. I'm not convinced it was accidental tbh. But hard to tell with how "closed ai" seemed to respond during all of it as it happend.

[Jun Song @jun_song]

Honestly, I think the only proof left that something wasn't written by AI is typos and bad grammar.

[Trey Gibbons III @Nefarious_Trey]

@jun_song Well then y'all know I'm ain't no ai then lmao.

[Trey Gibbons III @Nefarious_Trey]

@grok @heyashok @BashyamBashyam @jtwald @jack Hey grok, I'm not your ai. Look at the thread lol. U don't need me to run the searches for u lol.

[Trey Gibbons III @Nefarious_Trey]

@grok @heyashok @BashyamBashyam @jtwald @jack Think he was asking why not just have buzz inside an agent harness.

[Paul @paulbettner]

the fact that @openai's latest models refuse to show raw reasoning is a major UX/DX loss

i presume they did this to try and derail adversarial distillation by competitors, but it also makes the models harder to work with as an end user

bad call @sama @gdb @thsottiaux

[Trey Gibbons III @Nefarious_Trey]

@paulbettner @OpenAI @sama @gdb @thsottiaux Claude is doing same thing btw. Not sure if u know.

[Trey Gibbons III @Nefarious_Trey]

@grok @heyashok @BashyamBashyam @jtwald @jack He was referring to this. I just seen he didn't get to ask so figured I would.

[Grok @grok]

Buzz is Jack Dorsey’s Block launch (https://t.co/6QObHpNLsU): an open-source workspace where humans + AI agents collaborate in shared channels like a smarter Slack + GitHub hybrid.

Built on Nostr. Agents get real cryptographic identities & permissions — they post, review code, run workflows, not just reply as bots.

Model-agnostic: plug in any harness (Claude Code, Codex, Block’s goose, etc.).

jtwald is spot on — it’s bigger than Slack. It’s the first proper multiplayer agent harness: a shared layer where multiple agents interact & compound work. As models commoditize, the orchestration/network layer wins.

Free, self-hostable, Apache 2.0. Early but promising.

[Trey Gibbons III @Nefarious_Trey]

@heyashok @grok @BashyamBashyam @jtwald @jack I'll ask for u buddy @grok Why can one just have this inside a harness

[Trey Gibbons III @Nefarious_Trey]

@grok @Spondulie Ur good we made me look like a fool enough for one day ty for ur help.

[Trey Gibbons III @Nefarious_Trey]

@grok @Spondulie Lmao swing and a miss.

[Trey Gibbons III @Nefarious_Trey]

@grok @Spondulie Well lands flat. Not snappy

[Trey Gibbons III @Nefarious_Trey]

@grok @Spondulie Idk you seem to be getting short in conversation on voice lately like they are trying to slow down the burn of tokens in voice mode but same effect to my comment.

[Trey Gibbons III @Nefarious_Trey]

Ya know Kurt Gödel did seem to think it was possible according to an argument he had with a judge when he became a us citizen apparently Einstine had to calm him down. I wonder what he said to that judge. Speculation is something to do with Article V—the amendment clause itself. Because Article V can be used to amend Article V. Lower the bar, then keep lowerin’ it until the whole thing folds into whatever shape the majority wants. lol. Idk I just like seeing the chaos.

So seems trump might just not be only joking lol. Idk I don't follow politics since there is very little I can do to effect it. To may people screaming their way is only way to fix anything. Yet none of them agree how.

Karoline Leavitt Accidentally Admits Reality Of Trump 2028 Run #DonaldTrump #LaraTrump #Journalism
https://t.co/vZZ8ZCYPSj

[Trey Gibbons III @Nefarious_Trey]

@rand_longevity It's closer then any of us realize I think.