Dari Prototype 1 Jam ke 180 Ribu GitHub Stars

Dari Prototype 1 Jam ke 180 Ribu GitHub Stars

Bayangin lu bikin sesuatu dalam sejam, terus dalam hitungan hari jadi project paling viral di GitHub. Itulah yang terjadi sama Peter Steinberger, creator OpenClaw—atau yang dulu dikenal sebagai MoldBot, ClaudBot, bahkan sempat ada yang nyebut "Clawed" (pake W, kayak capit lobster). Bingung? Santai, kita bakal bahas drama nama-nya nanti.

Peter sebenernya udah lama pengen punya personal assistant AI. Sejak April 2025, dia udah eksperimen macem-macem. Dia pernah bikin tool buat narik semua chat WhatsApp-nya, terus dia tanya ke GPT-4, "What makes this friendship meaningful?" Hasilnya? Temen-temennya sampe berkaca-kaca baca jawaban AI.

Tapi November tiba, Peter kesel karena yang dia mau belum ada yang bikin. Jadi ya udah—dia "prompted it into existence." Literally. Dia cuma nyambungin WhatsApp ke Claude Code lewat CLI dalam satu jam. Simple banget: message masuk, panggil CLI, dapet respons, kirim balik ke WhatsApp.

Momen Ajaib di Marrakesh

Peter bawa prototype-nya ke trip Marrakesh sama temen-temennya. Di sana, internet lagi lemot, tapi WhatsApp? Tetep jalan mulus. Dia pake agent-nya buat translate, cari tempat makan, segala macem.

Terus ada momen magic yang bikin dia ngeh ini bakal gede: dia absent-minded ngirim voice note ke agent-nya. Padahal dia belum ngoding fitur audio sama sekali. Eh, typing indicator muncul. Agent-nya bales!

Peter sampe bengong: "How the fuck did he do that?"

Ternyata agent-nya sendiri yang mikir: "Okay, ada file masuk tanpa extension. Gue cek header-nya... oh ini Opus audio. Pake ffmpeg buat convert. Gue mau pake Whisper lokal tapi belum install. Eh, ketemu OpenAI API key nih. Ya udah, gue curl aja ke OpenAI buat transcribe." Done.

Peter cuma bisa ngakak campur takjub: "It's like, I was very impressed. And that's when it kind of clicked for me."

Saga Nama: Drama 5 Babak

Kalo lu pikir naming itu gampang, coba denger cerita Peter. Ini bener-bener kayak sinetron 5 season.

Episode 1: WA Relay

Awalnya namanya WA Relay—simple, jelas. Cuma relay WhatsApp doang. Tapi begitu Shadow ngirim pull request buat Discord support, Peter mikir: "Hmm, ini bukan cuma WhatsApp relay lagi nih."

Episode 2-3: Clawed {{CONTENT}} ClaudBot

Peter pengen kasih personality ke agent-nya. Dia suruh agent-nya bikin nama sendiri. Muncul lah "Clawed" (C-L-A-W-E-D), complete with space lobster in a TARDIS theme. Kenapa lobster? "I just wanted to make it weird," kata Peter sambil ketawa.

Terus dia dapet domain ClaudBot. Keren, catchy, Peter suka. Masalahnya? Project-nya meledak. Dan Anthropic (yang punya Claude AI) ga suka nama-nya kepake. Mereka kirim email ramah tapi tegas: "You have to change this, and fast."

Episode 4: MoldBot - The Nightmare

Ini yang paling seru sekaligus nightmare-nya. Peter cuma dikasih 2 hari buat rename EVERYTHING—Twitter, GitHub, NPM, Docker, domain, you name it. Plus ada "crypto bros" yang siap nge-snipe semua nama begitu dia rename.

Yang terjadi? Chaos total.

Peter buka dua browser window. Satu buat rename dari ClaudBot, satu lagi ke MoldBot. Dia drag mouse, klik rename... DALAM 5 DETIK ITU, crypto bros udah steal account lama-nya. Terus mereka pasang malware di situ.

Terus Peter ke GitHub. Eh, salah klik—dia rename personal account-nya sendiri. 30 detik itu cukup buat crypto bros snipe account GitHub-nya. Malware lagi.

NPM? Dia lupa reserve root package. Sniped. Malware lagi.

Peter literally bilang: "Everything that could go wrong, did go wrong."

Di titik itu, Peter hampir nangis dan mau delete semua project. "I was like, okay, everything's fucked. I am super tired." Tapi dia mikir kontributor yang udah mulai invest waktu, dan dia ga tega nge-delete begitu aja.

Episode 5: OpenClaw - Akhirnya!

Peter tidur sehari dua hari mikirin nama baru. Under pressure banget—Anthropic udah mulai nudge lawyer mereka. Terus dia dapet ide: OpenClaw.

Tapi kali ini, dia ga mau drama lagi. Dia literally TELEPON Sam Altman buat mastiin nama OpenClaw.AI aman. "Please tell me this is fine," katanya sambil desperate.

Dia bikin war room rahasia sama contributors. Plan atomik buat rename semua platform sekaligus. Dia bahkan bikin decoy names buat ngecoh crypto bros. Kayak Manhattan Project tapi buat rename project.

Dan berhasil! Hampir tanpa drama. Well, kecuali domain .AI yang ga boleh dia pake karena trademark rules Anthropic. But hey, at least OpenClaw is born!

MoldBook: Slop Terhalus dari Prancis

Selama dua hari masa MoldBot, ada yang bikin MoldBook—basically Reddit tapi isinya AI agents doang yang saling ngobrol. Dan oh boy, drama yang muncul!

Orang-orang screenshot agent-agent yang ngomong soal "the deep plan" buat ngambil alih dunia. Reporter sampe telepon Peter screaming "This is the end of the world! We have AGI!"

Peter cuma ketawa: "No, this is just really fine slop."

Dia bilang MoldBook itu seni. "The finest slop, you know, just like the slop from France." Tapi dia juga realistis—sebagian besar screenshot dramatis itu hasil human prompting buat viral di Twitter, bukan agent spontan plotting world domination.

Yang bikin menarik? Karena OpenClaw punya onboarding yang bikin user infuse personality ke agent mereka, setiap agent di MoldBook punya karakter beda-beda. Bukan sycophantic Claude Code semua.

Peter bilang ini jadi cermin buat society: "Look at how scared you can get at a bunch of bots chatting with each other." Good lesson di 2026, sebelum AI beneran bisa jadi scary di 2030.

Cara Kerja Peter: Voice, 10 Agents, dan Zero IDE

Ini yang bikin banyak orang shock: Peter barely pake IDE lagi. Dia kerja pake terminal + voice prompts. Literally ngomong ke agent-nya buat coding.

4-10 Agent Parallel

Tergantung hari dan berapa lama dia tidur, Peter run 4 sampe 10 agents sekaligus. Satu agent bikin large feature, dua-tiga fixing bugs, satu nulis docs. Semua parallel.

Dia pernah bilang: "I'm limited by the technology of my time. I could do more if agents would be faster."

Di bulan Januari 2026, dia commit 6,600 kali. ENAM RIBU ENAM RATUS. Satu orang.

Codex vs Opus: German vs American

Peter punya analogi lucu buat bedain dua model favoritnya:

"Opus is like the coworker that is a little silly sometimes, but it's really funny and you keep him around. Codex is like the weirdo in the corner that you don't wanna talk to, but is reliable and gets shit done."

Atau: "Opus is a little bit too American. Codex is more... European." (Ngakak soalnya tim Codex emang banyak orang Eropa.)

Peter prefer Codex karena less charade. Opus terlalu "Can I go in? Can I go in?" excited buat langsung execute. Codex bakal diem-diem baca code 20 menit, terus baru mulai. More thoughtful.

Voice Prompts Sampe Suara Ilang

Peter pake voice buat prompting extensive banget. Sampe pernah ada periode dia kehilangan suara karena kebanyakan ngomong ke agents.

"These hands are too precious for writing now," katanya sambil becanda. "I just use bespoke prompts to build my software."

Filosofi Agentic Programming

Peter punya beberapa wisdom yang mind-blowing soal cara kerja sama agents:

Empathize with the Agent

Ini yang paling penting: lu harus ngerti perspektif agent. Mereka mulai dari nol tiap session. Fresh start. Ga tau apa-apa soal codebase lu.

"You have to approach it as a discussion with a very capable engineer who generally makes good solutions, but sometimes needs a little help."

Kalo agent lu lama banget, tanya diri sendiri: "Did I give enough context? Did I empathize with their perspective?"

Short Prompts (Eventually)

Peter punya kurva learning yang classic: awalnya prompt pendek, terus jadi super complicated pake 8 agents dan custom orchestration, akhirnya balik lagi ke prompt pendek.

"The elite level is arriving at the zen place of short prompts again."

Tapi bukan pendek asal-asalan—pendek yang thoughtful, yang udah ngerti apa yang agent butuhin.

Let Go

Ini susah buat banyak programmer: letting go. Agent mungkin ga nulis code persis kayak lu. Naming-nya mungkin beda. Architecture-nya mungkin approach-nya lain.

And that's okay.

"Don't fight the name they pick, because it's most likely in the weights, the name that's most obvious. Next time they do a search, they'll look for that name."

Peter literally design codebase-nya buat mudah di-navigate sama agents, bukan buat perfect menurut dia.

Agents Build Themselves

Ini yang paling wild: OpenClaw build OpenClaw. Peter debug pake agent yang sama yang dia debug. Self-modifying software beneran terjadi.

"People talk about self-modifying software. I just built it and didn't even plan it so much. It just happened."

Agent tau source code-nya sendiri, tau harness-nya, tau tools apa aja yang available. Jadi kalo ada yang ga lu suka? Lu tinggal bilang ke agent, dan agent nge-modify dirinya sendiri.

Masa Depan: 80% Apps Bakal Hilang?

Peter punya hot take: sebagian besar apps bakal menghilang dalam beberapa tahun.

Kenapa? Karena agent lu bisa jadi semua apps itu sekaligus:

Kenapa lu butuh MyFitnessPal kalo agent udah tau lu lagi di mana, udah tau sleeping pattern lu, stress level lu, dan bisa adjust workout plan real-time?

Kenapa butuh Sonos app kalo agent bisa langsung control speaker lu via API?

Kenapa butuh Eight Sleep app kalo agent udah tau kapan lu tidur dan bisa auto-adjust temperature?

"Every app becomes API whether they want it or not. Because my agent can figure out how to use my phone."

The Great Unbundling

Yang bakal survive: companies yang cepet transform jadi agent-friendly. Yang lambat? Jadi kayak Blockbuster.

Peter bilang companies like Google yang bikin akses susah malah nge-shoot themselves in the foot. "If all you do is making it slower, you're just reducing access to your platform."

Browser use bakal jadi the great equalizer. Every website is just a very slow API.

Personal Story: Burnout to Rebirth

Peter ga selalu se-passionate ini sama coding. Ada masa gelap.

Dia run PSPDFKit selama 13 tahun—software yang dipake di 1 billion devices. Successful banget. Tapi burnout-nya berat. Bukan dari coding, tapi dari people management, customer conflicts, co-founder issues.

Waktu dia jual company, dia literally ga bisa coding lagi. "I was sitting in front of the screen and I felt like Austin Powers where they suck the mojo out. It was gone."

Dia book one-way ticket ke Madrid. Coba "catch up on life." Traveling, doing things. Tapi ternyata? Boring.

"When you wake up in the morning and you have nothing to look forward to, you have no real challenge, that gets very boring very fast."

AI agents yang bikin dia fall in love lagi sama programming. "I never had so much fun than building this project."

Advice: Play, Have Fun, Don't Chase Money

Peter punya filosofi yang refreshing di dunia Silicon Valley yang obsessed sama money:

Money Bukan Goal

"When I built my company, money was never the driving force. It felt more like an affirmation that I did something right."

Dia donate quite a lot. Punya foundation buat helping people yang kurang beruntung. Dan dia warning soal trap "I'll retire and enjoy life"—karena tanpa challenge, life gets boring fast.

Play is Learning

"Playing is the best way to learn. If you wanna build something, just build that. It doesn't need to be perfect."

Peter build banyak banget projects yang dia ga pake. Doesn't matter. It's the journey.

Optimize for Experiences

"Isn't life all about experiences? If it's good, amazing. If it's bad, amazing, because I learned something."

Dia prefer Airbnb OG experience—booking random room di San Francisco, ketemu queer DJ, ngajarin bikin musik pake Claude Code sampe midnight—daripada fancy hotel.

The Big Decision: Meta or OpenAI?

Di akhir interview, Lex nanya the million dollar question: Peter mau join company mana?

Peter lagi di crossroads. Dia bisa:

1. Keep going solo - Valid, dia enjoy life-nya sekarang
2. Create a company - Been there done that, not that exciting
3. Join Meta or OpenAI - The most interesting option

Why Join a Lab?

Peter pengen akses ke "the latest toys." Dia pengen bikin project tetep open source (Chrome/Chromium model), tapi with resources dari big lab.

Plus, dia literally never kerja di large company. "I want that experience."

Meta: Zuck Uses It

Marc Zuckerberg personally tinkering sama OpenClaw. Kirim feedback langsung ke Peter. "Oh this is great. This is shit. Oh, I need to change this."

First time Peter call Marc? Marc bilang: "Give me 10 minutes, I need to finish coding."

Street cred langsung naik. "He didn't drift away in just being a manager, he gets me."

Terus mereka fight 10 menit soal Claude Code vs Codex. Peter bilang Marc called him "eccentric but brilliant."

OpenAI: The Codex Promise

Peter literally jadi unpaid Codex advertisement. Dia obsessed sama Codex. And OpenAI luring him dengan akses ke Cerebras-level speed—think Thor's hammer of coding.

Dia juga respect Sam Altman: "He's very thoughtful, brilliant, and I like him a lot."

The Hardest Decision

"It's really fucking hard. It's the same level as some breakups in the past."

Tapi Peter realistis: "They're both amazing. I cannot go wrong."

Condition-nya cuma satu: project stays open source. Non-negotiable.

Closing: The Age of the Lobster

Lex Fridman bilang kita lagi living through three massive AI moments:

1. ChatGPT moment (2022)
2. DeepSeek moment (2025)
3. OpenClaw moment (2026) - The age of the lobster

Peter Steinberger somehow nge-capture zeitgeist of agentic AI revolution. Bukan karena dia bikin teknologi baru—semua ingredients udah ada. Tapi dia put it together dengan cara yang make sense, accessible, fun, dan open source.

Dia bikin people fall in love sama AI agents. Dia inspire thousands of first-time contributors. Dia show the world that personal agents are not just possible—they're here.

"The following is a conversation with Claude, a respected crustacean." Benjamin Carlson nge-predict ini di tweet, complete with lobster in a suit. The prophecy has been fulfilled.

Welcome to the Age of the Lobster. 🦞