Dev Samaaj @devsamaaj

The math is not mathing. 🤨 Last year it was all about efficiency. Now AI budgets exist and token consumption itself is becoming a flex. I saw someone feed Opus 4.8 an entire Excel sheet just to generate a pretty PDF. Dude. There's a button for that. It's called Export as PDF. Free, instant, zero tokens. There's already a term for it: tokenmaxxing. The real problem is we're using our most capable models for everything instead of matching the task to the simplest reliable tool. Agents make it worse — a simple task becomes read → call → inspect → retry → validate → reread → fix → validate again. Token prices are falling, but if tokens per task keep exploding, cheaper tokens ≠ cheaper AI. Better question: what's the cheapest reliable way to solve this task? Sometimes it's an agent. Sometimes it's Python. Sometimes it's a smaller model. And sometimes, it's the damn Export as PDF button. 😂

@triviwritescode we are becoming very lazy and it's not too far when most of human brains become more lazy and will not do anything the gap of human intelligence will increase between humans as gap of poor and rich.

If you are building something in tech, drop the link👇

To kaise hai ap log Idea credit: @ravish_journo

@creative_nashra when you start writing on paper your thoughts will slow down.

There are many things that do not make sense, but top leaders in their domains only talk about them when it affects them directly.

Nithin Kamath @Nithin0dha

5 days ago

I think MDR on UPI was probably inevitable at some point, especially given how widespread UPI adoption has become. It could also lead to more competition, instead of just three apps accounting for more than 95% of the market. That being said, there are some use cases, like

654 1K 9K 1.6M 960

Indians feeling right now after UPI fee above ₹2000 transaction amount. Hum loot gye hai or lut jane do.

4.5 years of frontend, never really wrote a class. React runs on hooks and composition and you can skip OOPs entirely. Started Python for FDE this month. Turns out classes model real things: users, transactions, state that changes. React never made me see that, not even once.

Can a tech guy get's 1000 likes?

How many are actual smartest?

Elon Musk @elonmusk

2 weeks ago

If you want to reach the smartest people in the world, advertise on 𝕏

11K 11K 140K 20.3M 4K

Hi smartest dev, I post content on AI & Tech Let's connect.

OpenCode reportedly crossed 120,000 GitHub stars in 2026. Worth asking why. Cursor's pitch is a better editor. Copilot's pitch is deeper IDE integration. Both bet you'll pick a home and stay there. OpenCode's pitch is different: don't pick a home. Plug in whatever model you're already paying for — GPT, Claude, something local, whatever ships next month — and run it from your own terminal, your own editor, your own setup. The model contract underneath now matters more than the editor on top. Free here means no subscription fee. You still pay per token for whichever model you bring — nobody's giving away inference for free. After three coding tools shut down or repriced this year, choosing your own model looks like insurance now.

I tried so hard got so far But in the end Hook content works not informative 😔

If AI disappear today I will get my job back 🙂

More than a 1-in-10 chance AI wipes out humanity. That's the number an Anthropic alignment lead put out this week, punctuated with an exclamation mark, days after another researcher quit the company warning it was moving recklessly toward self-improving systems. The internet had its usual reaction: is this science, marketing, or both? Here's my problem with it. Every time this debate resurfaces, it swallows the room. AGI, superintelligence, extinction risk — heavy words that make everything else sound small by comparison. Meanwhile the actual, present-tense damage keeps getting starved of oxygen: → Junior dev roles quietly disappearing as agents absorb the work that used to train people up → Data centers pulling power and water at a scale most cities weren't built for → Support, content, and QA teams getting hollowed out with zero regulatory conversation happening None of that needs a percentage attached to be worth taking seriously. It's already measurable. It's already happening to people who don't have a podcast or an S-1 filing to make their case. I get why the doom framing is seductive. For researchers, believing you've built the most consequential thing in history is a very human temptation. And for companies about to go public, sounding scary plays well with investors who reward capability signals over caution. But it lets everyone off the hook for the boring, fixable stuff happening right now. You don't need a 10-year extinction timeline to justify better labor transition policy, transparent compute reporting, or basic guardrails on what agents can touch unsupervised. Extinction risk might be real — I'm not qualified to rule it out. But it's also the easiest risk to talk about, because nobody's accountable for it yet. The harms already showing up in job boards and power grids have names attached. That's probably why they get less airtime.

If you are not in tech let's connect

Happy Engineer's Day Let's connect guys.