Rahmat Wibowo compared five SEA websites via PageSpeed benchmarks while defamatorily branding rival founders as frauds and degree-fakers, accusing Abil Sudarman, Marchel Shevchenko, and Muhammad Alif Ramadhan of falsifying credentials and fooling institutions, all to elevate his own InfraLoka.

Five sites, one question: did anyone this, or did it just vibe into existence? Five sites, one question: did anyone engineer this, or did it just vibe into existence? Use Case of InfraLoka VS Vibe Coded & Piece Of Art fi, Rahmat Wibowo | rr, | + Follow ) We ran identical PageSpeed Insights captures mobile and desktop, same morning, same throttling profile across five live SEA-facing products, then cross-referenced the stack with Wappalyzer and the hosting footprint with IP geolocation. The scores don't measure taste. They measure whether discipline was applied after the prototype felt good. 5 19 How the capture was run Every site below was captured cold, on the same morning, with the same lab conditions no warm caches, no CDN priming. The Lighthouse ledger, in full Desktop figure shown large, mobile figure alongside. Core Web Vitals pulled from real Chrome UX Report field data where available. Infraloka.co.id By Rahmat Wibowo From InfraLoka Product of Institut Teknologi Bandung Engineering exceller ” for the cloud-native + The only site where desktop and mobile scores are essentially the same number. That's not luck it's a modern framework (Next.js 16 + Turbopack) doing the work a hand-rolled stack would skip. Assai.id By Abil Sudarman From ASSAI. , BINUS ONLINE Drop Out, Claim to have degree from University of London The Man Fooled KORIKA UNESCO and The Stimson Center i ® Identical lab LCP on desktop and mobile, but field INP more than doubles on real phones (88ms > 210ms) the gap between what Lighthouse simulates and what users actually feel when they tap. Mrscraper.com by Cahyo Subroto from MrScraper Part of Hampton , The Brand Theft TTFB climbs to 1.6s on mobile before a single pixel paints a server/edge problem, not a markup one. The 12- plus-technology stack shows real product investment (analytics, uptime monitoring, CAPTCHA), but none of it touches response latency. Data-sorcerers.com By Marchel Shevchenko From Data Sorcerers product of Universitas Teknologi Yogyakarta , The fraud who claimed to have a degree from Massachusetts Institute of Technology The one site that scores worse on desktop than mobile a sign the desktop experience is carrying extra weight (probably imagery) the mobile build doesn't load. It's also the only site that fails Agentic Browsing outright, meaning Al agents and crawlers may not reliably parse the page. Perfect10.id By Muhammad Alif Ramadhan From PERFECT10 Al and IMPHNEN Product of University of Singaperbangsa Karawang , The Man Who Claim to Have 5 Years Of Experience Building Production Grade Turns Out Just Piece of “4, The most evenly mediocre result on the board nothing fails badly, nothing passes cleanly. One LCP optimization pass would likely flip Core Web Vitals from failed to passed on both devices. Every metric, one table Mobile values shown throughout the harsher, more representative condition for most real traffic. Performance, Accessibility, Best Practices and SEO, averaged across desktop and mobile captures a proxy for how much engineering happened after the demo worked. Framework choice is the floor The two best-performing sites both run on modern, opinionated frameworks (Next.js / a managed marketing stack). Hand-assembled stacks topped out 15- 35 points lower on mobile Performance Desktop scores lie by omission Every site except InfraLoka shows a meaningful desktop-to-mobile drop. If a demo is only ever shown on a laptop, the real number is hidden until launch day. Agentic Browsing is the new blind spot Three of five sites fail at least one Agentic Browsing check the category teams are least likely to have tested for, and the one that determines whether Al tools can use the site at all. What "vibe coding" actually costs A platform that demoed beautifully, audited badly e Rahmat Wibowo (© 5 securty unrbios Ne sthenicaton (AP completely open) Ne mantengsletng fan detect see) processes] mets e puke pata wit sent ose eis secu sues sp tough) Montoring (eaducton isos go urdstectd) Documentation enone acted in heads) Lighthouse only sees the surface render speed, markup hygiene, metadata. It can't see what's running underneath. A. separate security and SDLC review of a public-facing procurement-audit platform, handling sensitive government data, found a different kind of debt entirely: the kind that doesn't show up in a performance score until it's exploited. The product itself wasn't the problem. The team had clearly built something smart strong data analysis, working algorithms, a platform that did what it promised. What was missing was everything that turns a working prototype into a production system. Open APIs with no authentication layer, reachable by anyone CORS configured to accept requests from any origin No rate limiting wide open to scraping and denial-of-service Zero automated test coverage across the codebase No CI/CD pipeline; deployments handled by hand No monitoring or alerting to catch issues in production No documented error-handling strategy Plain JavaScript throughout no type safety net This is the actual shape of the vibe- coding tax: not a slow page, but an SDLC maturity score that lands near 9% against a professional standard of 70%+. Speed of build and survivability in production are different axes entirely and only one of them shows up ina demo. #WebPerformance #SoftwareEngineering #SRE #Cloudinfrastructure #VibeCoding #SDLC #CyberSecurity #TechAudit #Alagents #Indonesia #StartupTech #EngineeringExcellence #DevOps. #CoreWebVitals #infraloka