Abangan Tech

@idextratime Dan setelah juara Pep langsung mengumumkan kepergian nya

🗣️ Mikel Arteta mengenang momen Arsenal juara Premier League... "Salah satu perasaan terbaik dalam hidup saya. Saya pulang ke rumah, barbeque di luar. Bahkan gak sanggup nonton pertandingannya. Saya cuma dengar suara-suara dari dalam rumah.. lalu anak saya berlari sambil nangis dan bilang: 'Kita juara, Daddy.'" 😭

@idextratime Di era AI dan analisis data canggih, Southampton pilih cara paling primitif.

🚨 TERKINI: FA resmi membuka investigasi terhadap Southampton terkait skandal "Spygate". Southampton telah dikeluarkan dari playoff Championship dan kena pengurangan 4 poin musim depan, tanpa bisa banding lagi. FA kini menyelidiki sejumlah individu, termasuk pelatih mereka, Tonda Eckert, yang juga terancam dipecat oleh klub. Sanksi tambahan, termasuk larangan beraktivitas, masih bisa dijatuhkan FA. ⚖️ 📝 @JacobsBen

My personal take after running all three at scale: 🥇 Claude Sonnet best quality-to-cost for complex reasoning tasks. Most consistent output structure. 🥈 GPT-4o best ecosystem, most third-party integrations, fastest for simple tasks. 🥉 Gemini 1.5 Pro best for long-context document tasks. Cheapest raw token price. But quality inconsistency costs you at scale. The honest answer: run your actual use case on all three for a week before you commit.

How to actually compare LLMs at scale: Don't ask: "What's the price per token?" Ask: - How many calls does it take to complete my task reliably? - What's the average latency at my expected load? - What does a 1-hour outage cost my business? - Can I optimize my prompts to cut token usage by 40%? - What's the engineering cost of switching models later? Total cost of ownership beats price per token every time.

Most AI startups pick their LLM based on benchmarks. Then their API bill hits $8,000/month and they panic. The price per token is the smallest part of the cost. Here's everything they should have compared first. A Thread

The developers who will thrive aren't the ones who fear AI. And they're not the ones who blindly trust it. They're the ones who treat it like a powerful tool that requires skill to use well. AI is the best multiplier in the history of software development. Make sure it's multiplying something worth multiplying. RT if you agree. Reply if you don't, I read every one.

How to be on the right side of this: > Always read the code AI writes before you ship it > When AI gives you an answer, ask "why?" and verify it > Use AI to go faster on things you already know > Use AI as a teacher on things you don't, then learn them > Treat every AI output as a first draft, not a final answer The goal isn't to use AI less. It's to use it with enough understanding that you'd survive without it.

Hot take: AI won't replace developers. Bad AI users will. I know that's uncomfortable. Let me explain why and what separates the two. A Thread

@HalaBarca19 @idextratime Yang lagi cedera bang wkw

@abangantech @idextratime Siapa bang?

🗣️ Antony mengaku tetap bangga dengan performanya musim ini meski tidak mendapat panggilan ke Timnas Brasil. “Saya tentu ingin berada di sana bersama tim nasional, tetapi saya tetap sangat bangga dengan apa yang sudah saya lakukan.” 🇧🇷😊

@idextratime @diarioas Real Madrid musim ini dapat 10 penalti di Liga, dan 0 yang diberikan ke lawannya, tapi tetap ngerasa paling dizolimi wasit. dan sekarang malah yang dilaporin wkwk

🚨 TERKINI: Asosiasi Wasit Spanyol resmi laporkan Real Madrid! Florentino Pérez dan Real Madrid TV dituding terus menekan para wasit lewat kritik berulang dan kampanye media yang dianggap memengaruhi jalannya pertandingan. 😳 📝 @diarioas