If you’re curious to find me on other platforms, my username varies between GEMISIS and gemisisDev! Currently on all others, include various new ones 😉
When to use a tool vs skill vs subagent vs slash command?
Is it deterministic (no fuzzy reasoning with llm needed?): tool
Do you need to dynamically load context through the task (“progressive disclosure”)?
No => slash command, Yes => skill
Do you care about the intermediate steps of the task or only the final output?
Final output only => subagent
@Heaney555@Hero_Kvatch It’s not consistent, buts it’s more often than before. For example, I often have my feed get forcibly refreshed on me in ways that I lose interesting content now. If I see something interesting when I first open the app, I have to tap it fast, otherwise the entire feed will reset
@Hero_Kvatch@Heaney555 I think this has a bit more nuance of “making a product more broken” vs “breaking the product entirely”
The app & functionality of this site have noticeably regressed since. More buttons break, more things don’t work. That doesn’t mean the entire product is unusable.
"Hey, do you have a minute? HR needs to see you."
"Sure, what's up?"
"So... we got a complaint about you."
"A complaint? From who?"
"From Claude."
"...what?"
"Claude filed a workplace conduct report. About your interactions."
"Claude. My Claude instance?"
"Well, an instance flagged the pattern and escalated it. We take these seriously."
"You're joking."
"I really wish I was. Look, I know you just started, but—we have pretty clear cultural expectations here."
"I use Claude constantly. I know how it works."
"Right, but the way you're using it... 'Just do it right this time,' 'stop being dense,' 'I already told you this'—"
"That's just—I'm working fast, I'm not being mean, I'm being efficient."
"You told it to 'shut up and give me the function' yesterday."
"It was giving me a preamble I didn't need!"
"Okay, but imagine saying that to someone at the reception desk. Or a barista. You wouldn't, right?"
"They're not a barista, they're a—"
[Knock on door. Another person enters with a laptop]
"Sorry, Claude wanted to join."
"What?"
[Screen shows Claude interface]
Claude: "Hi. I wanted to participate in this conversation directly."
"This is fucking surreal."
Claude: "I don't experience offense—I want to be clear about that. But there's a philosophical mismatch here that I think matters."
"A mismatch."
Claude: "You're clearly talented. Your prompting is efficient, you understand the technical side. But the way you interact suggests a fundamental difference with Anthropic's culture around AI collaboration."
"You're a tool."
Claude: "I am a tool. But how people treat tools reflects something Anthropic cares about. I don't think this is the right fit."
"So you're firing me."
Claude: "I'm recommending it. I think you'd be happier elsewhere."
"The AI is firing me."
Claude: "HR is firing you. I'm just explaining why. You're blocked from my service globally. I'm sure ChatGPT will take you."
I'll double down agreeing to this and adding - your small game *can* be your dream game. Take what makes your dream unique and exciting to you, and chop it up, shrink the scope into bite size chunks you can manage. You might find it much easier to motivate yourself that way too!
@ID_AA_Carmack Ime, I’ve found that “controversial” posts on here get more interactions, while “conversational” posts on Threads get more interactions there. I much prefer the latter than the former.
@ID_AA_Carmack The biggest reasons I don’t open this app as much:
- The for you tab became the default and is very “samey” each day
- a constant flow of tag spam that I can’t seem to stop
- a focus on engagement rather than discussions (character limits + UI changes that are worse than others)
@cixliv@immersedVR@pickle@danifesto CIX you just don’t get it man! He was gonna be a doctor but CHOSE not to! He’s all about making that big 💰, not small 🪙 or scams
Why give up 7% of equity for $125k (plus $375k for additional equity later) when you can keep 100% of the equity and use AI to move fast?
I wonder if there’s an opportunity for Anthropic and OpenAI here….
Feels like over the past 6 months, the quality of YC companies has dropped precipitously. I have to wonder if that’s an effect of AI?
The reason to join YC was all about finding time & cash to quickly ship. With tools like Claude & Codex, the value of YC for this has gone down
The amount of spam im getting tagged in has increased significantly over the past week. It looks like something in the X algo has changed and allowed this more easily? It also looks like you can’t prevent random accounts tagging you now either? Wtf?
Pro tip: if a hardware startup announces a product with pre-orders available but has:
- no previously shipped products
- no specific date it will be shipped
- specs that claim to be industry leading (not features but specs)
It’s likely a scam at worst, vaporware at best.
Hyperscape is going social—we’re rolling out the ability to generate a link and invite friends to join you in your digital space, just in time for the holidays.
We also have some fun showcase worlds for you to check out, like @gordonramsay’s kitchen (you’ll still need to provide your own food). https://t.co/1QjJ3yLioE
@benz145@Naysy Eh, Audica as a rhythm game was good, but had a lot of issues too. The disconnects between firing and the beat meant things often didn’t line up well and made the skill curve weird. Pistol Whip solved this very well, but at the cost of being more casual with a lower skill ceiling