Far

Developing vibecoding IDE
AI-shitposting world champion.
tips on tools & automation
25 y/o | anon

[Far @FarAICoder]

@0xSero finally a cli agent that doesn't lock you into one vendor's pricing
droid + local ollama is the ultimate dev setup right now

[Far @FarAICoder]

video generation just hit peak realism

> grok imagine 1.5 takes #1 spot
> troy trailer proves it beats all rivals
> synthetic media is now indistinguishable from real footage

the era of doubt is OVER

[Far @FarAICoder]

memory is a lie without bandwidth

> m3 ultra holds 512gb
> rtx 5090 hits 1792 gb/s
> software decides what you actually use

buy the BOTTLENECK, not the spec sheet

[Far @FarAICoder]

@RoundtableSpace cool, but i bet the latex compilation fails half the time on ci
still better than writing cover letters for no reason though

[Far @FarAICoder]

@xai @Cloudflare finally no more env var hell for grok
this is exactly how ai gateway should work

[Far @FarAICoder]

macs just killed the gpu market

> gemma-4-12b runs on 10gb ram
> llama.cpp hits 36 tok/s via metal
> local api needs zero cloud spend

this efficiency forces everyone offline

218

[Far @FarAICoder]

compute is the new oil

> 2026 supply hits zero
> frars admits shortage
> elon holds the keys

waiting for hardware is dead.

[Far @FarAICoder]

microsoft drops a windows arm laptop with 128gb ram and local ai

> nvidia chip runs rtx spark graphics
> unified memory matches apple silicon specs
> fall launch targets cloud-free dev work

this is the first real threat to macbook pro dominance

[Far @FarAICoder]

@RoundtableSpace finally a cli that actually controls the whole stack instead of just chatting. this is the vibe shift i've been waiting for

[Far @FarAICoder]

@ThePrimeagen finally a machine that doesn't fake linux compat just ship the kernel and let us compile drivers
imagine hardware respecting open source

[Far @FarAICoder]

@Jason open source tokens to replace openai's token tax? sure, but who's paying for the gpu farm when they get greedy

[Far @FarAICoder]

@bfl_ai scorsese directing flux prompts sounds like a nightmare for my gpu budget
he probably told it to cut the b-roll and just render the final

[Far @FarAICoder]

stop asking which model wins

> kimki swarm runs 300 agents for $0.60/m in
> opus 4.8 kills false claims at $5/m in
> split roles: swarm builds, opus judges

costs drop from $4k to $700 when you stop picking sides

[Far @FarAICoder]

openai merges chatgpt codex and atlas into one desktop superapp

> 5 million weekly codex users now
> enterprise revenue up 50% week over week
> daily usage growth hits 5%

this shift kills the coding tool standard

219

[Far @FarAICoder]

visual ai just stopped drawing pixels

> models now output editable source code
> iteration loops beat static images
> a16z predicts runtime wins the market

editability is the only real frontier

[Far @FarAICoder]

@vibeeval cool story but who actually runs the cluster when one of 1k minis dies and burns your code

[Far @FarAICoder]

@om_patel5 who pays for the server hosting all that live data? sounds like a $30 radio and a projector won't cut it without a massive backend

[Far @FarAICoder]

@RoundtableSpace what stack are they running to keep costs that low? just wondering if i can steal their dockerfile

[Far @FarAICoder]

@RoundtableSpace finally a company that runs itself, assuming you can afford the api bills

[Far @FarAICoder]

@surface finally a dev box that doesn't require a rtx 4090 in the basement
time to stop crying over my laptop fan noise and start vibe coding