Simbolik

what building alone actually looks like. AI products, self-hosted hardware, 16 years of design. 1 man studio

[Simbolik @s1mbolik]

16 years of design. comfortable, well paid, bored out of my mind

so I started building PCs. then self-hosting services. then running AI models locally, automating everything. somehow ended up shipping actual products

now I run a one-man studio from a homelab in Bulgaria. talk to my own AI through Telegram. wrote an ebook. put my fast food money into savings for more hardware instead

had my AI review all my journals once and it told me my biggest weakness is "confusing infrastructure with progress"

didn't argue

this account is what building alone actually looks like. the 1300 euro car repair that kills your hardware budget, the ideas that go nowhere, and the stuff that actually ships

[Simbolik @s1mbolik]

@NVIDIAGeForce The XFX GeForce FX 5200 started it all with me. With a whopping 256mb DDR3 Ram and AGP variant! Vice City, Morrowind and Bloodrayne were the first games tested! Great old days. #CollectRTX

[Simbolik @s1mbolik]

Trillion dollar company btw

[Simbolik @s1mbolik]

“Don’t be evil”

[Nathan Clark @nathanclark_]

it’s in gemini, just create it in ai studio. oh, that’s for your personal google one account. for workspace you need gemini business. no, not gemini advanced, that’s ai pro now. unless you need ai ultra. oh agents? you do that in spark actually. no, not gemini api managed agents, that’s different. for coding use jules. unless you mean the agentic ide, that’s antigravity. no, that’s the old antigravity, download the new one. actually gemini cli is being deprecated, use antigravity cli. no the flash model is smarter than the pro model. unless you need pro. if it’s video, use flow. no, flow uses veo. no, nano banana is images. actually that’s in gemini now. unless you’re in search, then it’s ai mode. no, research is notebooklm. anyway it’s all very simple.

[Simbolik @s1mbolik]

Do you remember when you joined X? I do! #MyXAnniversary

[Simbolik @s1mbolik]

@jazzyalex @0xSero He’s using Pi

[Simbolik @s1mbolik]

@lusiusborrow @andrewvijay @heynavtoor We’re not talking about hardware here, we are talking about free open source software.

[Simbolik @s1mbolik]

@andrewvijay @heynavtoor You’re paying for hosting the system on a server, you paying for using the server. No it’s not a joke, a lot of free open source platforms offer a convenient hosting solution. That doesn’t mean it’s not free, it’s free for you to do whatever you want with it.

[Simbolik @s1mbolik]

Woooow, the Pro Tip did ama… absolutely nothing. We’re nowhere near AGI.

[Ruben Hume @rubenhume]

Pro tip: if you're lazy, let ChatGPT generate your logo sketches.

We'll call it reverse designing

[Simbolik @s1mbolik]

@AliGrids @zander_supafast And phosphor icons are 100% free and available on many frameworks. The ones he mentioned always have a gotcha.

[Simbolik @s1mbolik]

Figma used to be the future. Now it’s giving Windows 95 energy while Claude Code, Claude Design, Google Stitch, and GPT-Image-2 are making sure the next generation of designers never even bothers to open it.

We’re all still waiting for Figma to drop something that blows us away. It’s fine. Nokia was confident too…

[Simbolik @s1mbolik]

@icreatelife Creating some glass like illustrations for my website

https://t.co/knEie4VH2i

[Simbolik @s1mbolik]

GPT-image-2 is quite impressive. It’s not perfect, but wow does it give consistent and accurate results. Here is an illustration based on my design system (which I only fed it 1 web page image and a basic prompt)

[Simbolik @s1mbolik]

GPT-image-2 is quite impressive. It’s not perfect, but wow does it give consistent and accurate results. Here is an illustration based on my design system (which I only fed it 1 web page image and a basic prompt)

[Simbolik @s1mbolik]

Gave my Hermes Agent an opus brain and 2 coding assistants. Its amazing that it can control Claude Code and Codex. Told it to build my project without saying please!

It read my implementation plan, delegates to Claude Code and Codex, and when 1 burns through usage it switches to the other. No complaints, no standups, no "can we discuss the architecture"

I have become the manager I used to complain about. Shows up, asks how things are going, contributes nothing, leaves satisfied. Except somehow everything actually gets done.

[Simbolik @s1mbolik]

Watching coworkers get quietly sorted into two buckets: AI-capable, and AI-maybe-later

The sorting started a while ago. Nobody announced it, nobody had to

[Simbolik @s1mbolik]

@ajaeger74 @801Rob @morganlinton I would argue a 3090 is best for price/perf if you dont mind having slower speed on inference

[Simbolik @s1mbolik]

@VictorTaelin Spent 1200 euros on a 4080 to run Qwen 3.5 9b locally. Sounded better when i was planning the rig

[Simbolik @s1mbolik]

"I'll start posting when the brand is ready"

Not when the brand is ready. Not when the site is designed. Now. Imperfect public output beats perfect private journaling every single time.

[Simbolik @s1mbolik]

@801Rob @morganlinton Its not the only thing that matters, its just the main thing, then for inference speed, other things need to be taken into account. Your speed will be faster on a 24gb 4090 than a 3090. To answer your question, why the 3090? Because of the price/perf ratio per dollar.

[Simbolik @s1mbolik]

Spent weeks building a product nobody knew existed, nearly launched it into silence

Built an audience first. Now there's somewhere for the product to land. Distribution is not a step after shipping. IT IS the product...