Nik

@beyoumf Unplug everything. Buy a Nokia 3100, have 200 minutes of phone talk and 50 txt limit. Only people you love have that number. Spend the rest of your days face to face with anyone that matters to you. Material desire is the ultimate tar pit.

@Suupercharged When a new AI product drops, I just automatically assume it's all hype. 9/10 times, it usually is.

Tried Claude Design today and don’t get it. It gave the same output as before, but at 5x times the usage rate. What am I missing?

@BusMark_w_Nika You're absolutely right!

Nothing makes you humbler than the code that doesn't work.

@RGA why is there a need for a startup for this?

everything is stupidd

@bruvimtired React was a mistake.

useEffect was a mistake. hooks were a mistake.

@arlanr The cost of LLMs getting out of control, they are only viable for if you can offset the cost completely to run them. Local LLMs might be the solution for hobbyists.

what do people think about personal ai agents like openclaw? did anyone churn in the past few weeks or months after using them? also, do people pay for any cloud versions here?

@htmx_org thanks for reminding me how shitty the snow was this season in Colorado.

iykyk

@jshguo Most abstractions will be infra and users won't even notice the difference in their daily interactions. The chatbox is clearly a stop gap medium. Most users will want the AI working in the background not something they have to constantly orchestrate.

I have a feeling AI will fundamentally reshape software and interaction design. At the same time, “just add a chat box” doesn’t feel like the real answer, which is more like an early, rough phase. As for what it turns into…I genuinely don’t know yet.

@zerohedge It's your choice - Microsoft

*MICROSOFT TO OFFER VOLUNTARY RETIREMENT TO 7% OF US WORKFORCE

Based on how much I pay for LLMs every month, I'm gonna need for them to come in the office 3x a week, at least.

@htmx_org Thank you for your attention in this matter.

blocked the creator of fixi.js a LONG time ago, just a complete jerk and very low signal to noise no regrets

@adamc0dez Abstracting the work tells us that maybe writing that email wasn't worth the effort to start with. Mosts tasks are just noise, giving them to the AI doesn't really solve it.

AI replacing writing work emails is something I'm ok with

@elgermerlo Specs & planning. Until you have enough great examples of a good plan / prd / tasks, never hand that off the the agent. Agents need to see what 'good' looks like first.

Only for vibe-coders: what’s one thing you wouldn't delegate?

@damnGruz CODEX seems to be pretty good with a good plan.

i want to refactor a codebase, what should i use codex or claude?

@harshilmathur Planning for an entire year is basically a fantasy novel

If you want to read fiction, just read any tech company’s annual plan this year.

@GaelBreton i think people are using other models, I know I am. there comes a point where the other available models are good enough, where context, /skills and harness matter more.

Seeing less and less Opus 4.7 complaints. Did you adapt, quit or did it get better?

@PovilasKorop Not surprising to tbh. I've been using 5.4 for over a week now, it's comparable to Claude in many tasks. Also some platforms are forcing users to use 5.4 / 5.3 CODEX due to Claude limits.

Oh wow, that's unexpected from GPT-5.4. I guess so many people are switching from Claude Code, and we will start having identical problems on the other side now?..

@pugson It's just cheaper to hire devs at this point due to token burn. Every platform is pivoting due to insane costs. It will be interesting to say the least.

this next wave of tech layoffs is going to be crazy. every big company is desperately trying to patch up the hole in their budget caused by AI spend in Q1 after telling every employee to stop writing code.

@alexjpanagis Taking bets what's going to 'cook' designers in Q3-Q4

Ok so, nobody act surprised, designers are safe! Claude Design was overhyped once again... 😅

@aarondotdev Basically one well written skill at this point

all AI code review products can literally be replaced by claude code / codex and a single prompt

When LLM cost more than salaried Devs, do you then go back to hiring devs again?