Chetan

Built a $2 million ARR trading company & left everything to build a cultural enlightenment tool | systems orchestrator | NZ 🇳🇿

[Chetan @technaski]

Been running the same design prompts through GPT-5.6, Fable 5, and Grok 4.5 for the past few days and wasn't expecting the gap to be this obvious..

GPT-5.6 reasons like nothing else:
→ ask it why a layout works and you get an answer better than most actual designers could give you
→ ask it to just make something beautiful and it plays it safe
→ every output feels like it's showing its work instead of trusting its taste

Fable 5 can genuinely produce great design, that's not the issue:
→ feels less like using a tool and more like negotiating with it
→ you end up doing math before every prompt instead of just designing

Grok 4.5 just does the job:
→ prompt in, finished layout out
→ spacing, type, hierarchy, all of it already feels considered
→ saves you fifteen minutes usually lost nudging padding
→ it's sitting inside x premium, so there's basically no friction to start using it

The best model isn't the one that produces the prettiest screenshot. It's the one that gets you from idea to production with the least friction.

Right now that's grok 4.5, and it's not close.

[Chetan @technaski]

@levelsio telling claude to talk like an aircraft maintenance manual is the most levelsio solution ever

[Chetan @technaski]

the era of 100x cheaper than OpenAI or Anthropic has an expiration date

they built the brand on being the cheap option, now they have to prove they're the best option

[Chetan @technaski]

@levelsio google scraped the web for training data and now everyone's building private internets

[Chetan @technaski]

@levelsio portugal's the waiting room, the real builders already moved to asia fr

[Chetan @technaski]

@levelsio customer money is the only clean money, everything else has fine print

[Chetan @technaski]

@quxiaoyin Deepseek works really well alongside Hermes, yet to see how Deepseek Code performs

[Chetan @technaski]

vibecoders waiting for the 8th of August like

[Chetan @technaski]

@WatcherGuru This was probably for the best

[Chetan @technaski]

@levelsio launching a coffee brand on a random Wednesday is peak levels energy

[Chetan @technaski]

This is what pain looks/feels like:

[Chetan @technaski]

I could see model-agnostic harnesses becoming the next major wave in AI.

OpenClaw won the attention, but Hermes is winning on product depth: persistent memory, self-improving skills and the freedom to switch models without switching workflows.

DeepSeek entering with its own harness will reignite the competition, and when every model is forced to compete inside the same interface, the user wins.

[Chetan @technaski]

Kimi over delivered, Qwen just landed and DeepSeek is cooking

As an AI user, have you tried open source models yet? What did you think about them?

I run them alongside my Hermes Agent and the results have been nothing short of exceptional

[Chetan @technaski]

@naval the terminal was the api all along, agents just made it obvious

[Chetan @technaski]

Is this AI or not?

[Susan Abel @susylicious2023]

Reconnecting with my ex-fiancé’s child from the affair after 23 years

[Elon Musk @elonmusk]

@jk_rowling Thought that was Starship for a second

[Chetan @technaski]

The accounts that grow just by asking the silliest questions, will never cease to amaze me.

There's a market for stupidity, have you noticed?

[Chetan @technaski]

@botblastcap @imbabybrooklyn Either that or your model/provider is just slow, try again with Grok or Deepseek, should see an immediate difference

[Chetan @technaski]

AI has killed freelancing

I remember the days where I was spending upwards of 2000usd for a website that looks like what the Claude x Higgsfield stack does for a few bucks now

[Chetan @technaski]

@TheGeorgePu Was this Grok or Deepseek?

[Chetan @technaski]

The alt-Instagram crowd discovered Claude Code and never looked back, whilst Codex quietly became the better tool at everything except being performative