Mario Christian

Built a CLI tool that answers questions about your API spending in plain language. Fed it my logs, asked three questions. It told me I was overpaying, using Opus for tasks Sonnet could handle much cheaper. GitHub: github.com/mccode86/ai-co… #Python #AnthropicAPI #AIEngineering #LLM

@James_Buthuanal @vishaltweetup Exactly. Prompting isn't just syntax, it's a mental model you build over time with a specific model. That doesn't transfer easily.

@mc_code_ @vishaltweetup This is actually so underrated. The models behave differently and unless you're an ai researcher you just can't keep up with the pace of the development and u can't tailor your prompts and workflow to match the reasoning and thinking patterns.

As a developer, which one are you choosing? - GPT-5.5 - Claude Opus 4.8

@VaibhavLLMs Accuracy for me.

Be honest , as a developer , what matters most in an AI model ? - Cost - Latency - Accuracy - Efficiency 👇

@crypto_guro @jackprice Followed!

@mc_code_ @jackprice Hi mario, let's do it

Goodnight thought: in music, I learned the silence between notes matters as much as the notes. In code, the blank lines and spacing I used to ignore are what make it readable. Same lesson, different room.

@icanvardar Tools change. The obsession doesn't. That part was always the differentiator.

ai is already doing the repetitive tasks but it can never replace the human who decides to obsess over a vision and ship every single day. builders don’t fear technology they use it as fuel while everyone else is busy making excuses about how the world is changing too fast

@Tim_Denning Exactly. The dream isn't zero work. It's full signal, no noise.

The goal isn’t to never work again. It’s to never do bullsh*t work again.

@pmitu Totally agree Paul — as a human who definitely types every word manually with my ten organic fingers — this resonates deeply.

Don’t ruin your reputation with AI generated replies

@Jayyanginspires The sharpening is the work. A well-crafted prompt doesn't just save time, it forces you to understand exactly what you want. Most people skip that part and wonder why the output is off.

Give me 6 hours to complete a task, and I will spend the first 4 hours sharpening the prompt.

@pmitu Depends what you're building. For technical credibility and real-time conversations in AI, X wins. LinkedIn is where you go after you've already built the reputation.

LinkedIn connections > X followers

@TTrimoreau Because building the frontier is more interesting than monetizing it.

If anyone can make $1M a month using Anthropic Claude. Why are the engineers who built it still working at Claude ?

@btw_iamShreyaa Both, but for different reasons. Writing it manually builds the instinct to read and judge AI output. Without that, you can't tell when it's wrong. And it is wrong, sometimes confidently.

As Claude, Codex and other LLM's getting smarter every day Is learning to code manually still worth it? - Or is understanding the concepts enough now? 👀 Genuinely asking 👀 !!!!!

@zavxai And the cat doesn't even look stressed. That's the part that gets me lol 😂

Me: writes the first 10%. Claude Code: finishes the remaining 90% in 20 minutes.

@SakshiSugandhi Genuinely curious about the logic here. Starting with a weaker tool to build habits, then upgrading? Or something else? Because I'd argue learning with the best tool from day one sets a higher baseline for what "good" looks like.

Unpopular opinion: Learn debugging with Grok or Meta AI first. Then move to Claude or Codex. Thank me later.

@FrankSellhausen @paulg Always the outsiders. Too small to defend the old model, just hungry enough to build the new one.

@mc_code_ @paulg Spoiler alert: it’s typically not the large established companies that figure this out. It’s usually the small nimble companies that aren’t bogged down by bureaucracy.

If big companies can't make a net return on their LLM token costs, that doesn't mean it's impossible to. In fact this is exactly what you'd expect to happen with a new technology. Incumbents can't use it well, and are replaced by upstarts who can.

@github Where beats meet bytes. Studio desk counts, right? 😁

Not all great work happens at a desk. Where's your go-to spot? Show us 📸

@SahilBloom The hardest part of learning something new isn't absorbing the knowledge. It's unlearning the assumption that you already know how things work.

I'm convinced that adaptability is the highest form of intelligence. Knowledge matters, but the ability to learn and unlearn matters more. To change your mind in response to new information. The rigid cling to what was. The adaptable adjust to what is. The future belongs to them.

@GaryMarcus The gap between "we think a pause would be good" and "we are pausing" is where all the interesting corporate behavior lives. Not just Anthropic.

Funny how so many people read Anthropic is calling for a pause when they did NOT actually call for a pause. Read what they said, carefully. They want it both ways. They *don’t* actually want a pause - at least for now. Rather, they want to rush ahead, hinting at “least cautious actors” for justification. Instead, they want people to talk about an “option” they don’t actually plan to take, and are unlikely to ever take. (More likely, they will like always hint at China, and continue rush ahead.) It’s an incredible, cost-free piece of rhetoric — perfectly timed for the IPO.

@PeterDiamandis I'm learning AI engineering from scratch with no CS degree. Just Claude, consistency, and real projects. The credential gap is closing faster than universities want to admit.

Hot take: Universities charge $300,000 for a degree that teaches you skills any LLM can do for free. At some point we need to have an honest conversation about whether higher education is the greatest individuals misallocation of capital in recent history.

@PhysInHistory Wanting something. AI can optimize for a goal you give it, but it can't want anything on its own. That gap might be the most important one.

In what ways do you think AI will never surpass human intelligence? ✍️