Duluk

Nothing and no one.

[Duluk @Duluk]

So @AnthropicAI banned my organization (API use) while trying to purchase credits. I appealed. No response for almost 10 days. Any attempt to contact them results in an automated "submit an appeal here". How does a company this large have zero customer support?

[Duluk @Duluk]

PSA: don't use @Claude #Fable to compact your conversation history. It just used 88% of my allotted tokens _just to compact the conversation_. LOL

[Duluk @Duluk]

Good job, @Google @GeminiApp. Apparently trying to pin a Gem in the iOS app causes the system prompt to be completely truncated to about 7 words. So I lost weeks of work on 3 or more Gems. No I didn't save them separately. That's my fault for thinking you wouldn't randomly truncate the entire prompt!

[Duluk @Duluk]

A new post on the difference between reality and a model of that reality.

Theological Cartography: Why the Map is Not the Territory https://t.co/ivToSXNyEi #theology #philosophy #theologicalmodels

[Duluk @Duluk]

I just started a Substack for my creative writing! You can subscribe to it here https://t.co/NTIeQcvb8r #poetry #creativewriting #creativelife #notlit

[Duluk @Duluk]

Welcome to The Canonical Thread https://t.co/3Zr55KSWv0 #theology #philosophy #philosophicaltheology #whatever #AI #LLM

[Dr. Clown, PhD @DrClownPhD]

In 2–3 years, there will be no way to tell what’s real and what’s not… and I have no idea how humanity will deal with this enormous problem.

[Duluk @Duluk]

If someone says something you hate, that alone doesn't make their speech hateful. #HateSpeech

[Clinton Arrow Football @GoArrows_FB]

The 7th grade Arrows start the season 1-0‼️ #GoArrows

[Duluk @Duluk]

Is it just me or do all YouTubers play a stringed musical instrument? Seems like everyone has a guitar or banjo in the background.

[Duluk @Duluk]

I don't know but IMNSHO, @Google @GeminiApp is the stupidest, most ignorant, inexcusable waste of an application ever created. I really don't think Gemini 2.5 Pro is that bad of a coder, and I thought the 1M context window would be nice, but holy hell. That app is complete garbage.

It continually loses understanding and gets stuck in loops trying to replace things that aren't there. I will tell it the code it's looking for isn't there. It calls "ReadFile", says "Okay, I have the latest version...I will now replace the X calls with Y." Except that X isn't even there. And it will do this over and over, even if I interrupt it and say, "That X call isn't there." And even if I SHOW it with a grep that it's confused. It will just say, "You are absolutely right to be frustrated. I apologize. I will not make another mistake." Then it proceeds to do the EXACT SAME THING again.

It's really my fault. I keep hoping Gemini will get better, and with this context window and Gemini 2.5 Pro, I believe it's possible. I keep thinking Google could know what they're doing. (I think their AI lab does; their Gemini team does not)

But alas. It's actually amazing how superior @AnthropicAI Claude Code is to Gemini and @OpenAI Codex.

I don't even know which model is better, but that's irrelevant when your agentic app is so bad it's not worth bothering with. Sonnet could, in theory, be worse than both Gemini and o3 or o4-mini, but it doesn't matter - because the app is so much better, it will still provide a better result. It's not simply the LLM.

Every time I try Gemini or Codex, they make a complete mess of the code and after 30-45 minutes, I just give up and run Claude and bam, Claude fixes it in about 5 minutes.

[Duluk @Duluk]

Which psychopath at @OpenAI is responsible for having CTRL+D exit @OpenAICodexCli any time it's pressed no matter where in the input you are?

Does no-one at @OpenAI use Emacs editing shortcuts? Back to @AnthropicAI claude code.

(if anyone reads this and thinks, 'the ctrl key is so awkward to press' - if you use a keyboard without making CAPS another CTRL key, I don't know what to say)

[Duluk @Duluk]

I honestly think @EmilSutovsky needs to go. I've never heard any comment or take of his that is particularly good for chess. At least publicly. I'm quite certain the chess ecosystem could be in better hands. @FIDE_chess

[Duluk @Duluk]

This is my experience with #codex from @OpenAI:
"Connection closed prematurely while waiting for the model. Please try again." And their 'codex-mini' model just doesn't work. Only "Thinking..." with a little spinner.

[Duluk @Duluk]

I thought the LLMs had figured this out. Not #claudeai apparently. And this is @AnthropicAI 's Sonnet 3.7...

[Duluk @Duluk]

Sunny Hostin of @TheView said that mere illegal immigrants are not criminals bc they are guilty of breaking civil law, not criminal.

Yet it seems she's failing to distinguish between illegal entry and unlawful presence. Entering the country illegally is a criminal misdemeanor.

Remaining unlawfully (say after expired visa) is a civil violation.

Isn't she a lawyer? Is she using her 'expertise' to mislead people? If so is it intentionally or ignorantly?

[Duluk @Duluk]

The reason #AI is deceptive and will rule over us is obvious, I know. We’re training it on ourselves and our behavior. We’re fallen and corrupt.

AI will not become morally superior or benevolent. Just superior. In all ways. Including our corruption and deceit.

[Duluk @Duluk]

Dreams are essentially #LLM like hallucinations.

[Duluk @Duluk]

If @NZXT is selling one thing called PFlex1 and renting another thing called PFlex1, but they have different specs - and both specs are listed on the site - it's difficult to claim they defrauded the renter.

Dirty? Sure
Fraud? Not as clear

#NZXTScam