Edy

Late night coding is a poem

@hdd_edy nah thats senior dev coding not old man coding for sensitive algorithms managing real assets you want human brain on the logic, AI on the cleanup the 80/20 split sounds about right. i do similar - claude for refactoring and tests, but the core decisions stay manual

Most of my coding work involves complex, sensitive algorithms managing important assets. So I still write 80% of the code by hand, then tell Claude Code to “refactor where needed.” Am I coding like an old man? 😭

@eng_khairallah1 @eigencloud @buildoneigen @sreeramkannan @0xpratik Well deserved bro 💪💪💪

holy moly mommy i made it 😭 i have won two places in a row twice i have never participated in any program in my life because i did not see a project that suits me but eigencloud is different, and i have been saying this since 2024 thanks for all your support love you <3

if you fall for this propaganda you are genuinely ngmi

AI development videos be like: I will say stuff that any smart developer with common sense should know, while drawing on the screen to make it look techy and smart

Just remember guys are spending $1000s on Mac Minis to spin up an "AI Agent" to vibe code basic front ends with basic databases so they can pretend online that they have an AI software company that has $0 in revenue and 0 customers.

@BitcoinNewsCom How does he have the audacity to talk about this as if it's not another rope of slavery around their citizens' neck

NEW: Dutch Parliament Member Michel Hoogeveen explains how the 36% unrealized capital gains tax, just passed by the House of Representatives, will work. Here is a more detailed example: Step 1. Starting position You own 500 shares. Value on Jan 1, 2028: €50,000 Value on Jan 1, 2029: €100,000 So the paper gain is: €100,000 − €50,000 = €50,000 unrealized profit You did not sell. But for tax purposes, that €50,000 is treated as income. Step 2. Apply exemption You are married, so you get a €3,600 exemption. €50,000 − €3,600 = €46,400 taxable amount Tax rate: 36% €46,400 × 36% = €16,704 tax bill That bill is due in May, even though you never sold anything. Step 3. Market falls before you pay Now suppose by May the shares drop in value. New total value: €60,000 So your portfolio is no longer worth €100,000. It’s worth €60,000. But the tax bill is still €16,704, because it was calculated based on the January 1 valuation. Step 4. You must sell shares to pay tax To raise €16,704, you sell part of your shares. After paying the tax, you’re left with: €60,000 − €16,704 = €43,296 Originally you had 500 shares. Now you have 360 shares left. You were forced to sell 140 shares. 140 ÷ 500 = 28% of your shares gone. Step 5. What happened economically? Before the correction: Paper gain was €50,000. After the correction: Portfolio is worth €60,000. Original cost basis was €50,000. Real gain is only €10,000. But you paid €16,704 in tax. So instead of being up €10,000, you are now: €43,296 − €50,000 = €6,704 below your original starting value. You turned a €10,000 real gain into a €6,704 net loss. And you lost 28% of your shares permanently.

@DefenceHQ @BritishArmy @Leonardo_UK *I zpeak englis ser*

Soldiers in the @britisharmy will get new tech that can detect and locate enemy fire on the battlefield 5 years ahead of schedule. As part of a new contract with @Leonardo_UK, 250 UK jobs are sustained and 29 SMEs benefit in the supply chain. Read more👇 ow.ly/tJqa50Yfeik

@RoundtableSpace *This is humans button*

AI agents are now wondering if they dream in code or just reboot in silence. 🤣💀

@TheCryptoProfes I'd rather not trade

Would you rather: A: Trade funds in a wallet you don't own B: Trade funds in a wallet you do own

@0xKimberly Correct me if I'm wrong, but this is the damn close to Oman, which you have to drive 3-5 hours to reach...

The Dubai they don’t show you 🇦🇪

@aashatwt @chainyoda A web3 vibe coder is as catastrophic as it gets

if you are into crypto x ai x vibecoding join us here

@Jeremybtc Absolutely

Are prediction markets considered gambling?

@milesdeutscher Revenue what? They spend thousands of dollars in tokens if they are running 24/7, and there honestly isn't enough shit for them to do to make more money back

bro You literally CANT be lazy right now This is your competition HUNDREDS of AI agents working autonomously at once (thousands in revenue btw) Lock tf in

@Cointelegraph 'State-controlled surveillance app' as if whatsapp is not a US sate-controlled surveillance app

🇷🇺 NOW: Russia blocks WhatsApp messaging service, removing it from internet directories to push citizens toward state-controlled surveillance app.

@0xNairolf How would that happen if we're reliant on company run LLMs? Home run LLMs will never be as powerful as company run LLMs.

there is a massive market for making ai agents trustless

@whynotbit I don't really believe in trading, and I highly discourage it. I doubt even an AI can perform well in it, otherwise OpenAI and Anthropic would be swimming in trading revenue.

@hdd_edy Performing AI trading bot ?

I’ve been sharpening my AI agent-building skills over the past few months, and I’m now at the point where I can build genuinely sophisticated, high-impact agents… but I’m stuck on what to build next 😅 So I’m opening it up to you. What AI agent would be truly useful for you, your team, or your company right now? If it’s a real pain point and an agent like it doesn’t already exist (or existing tools don’t do it well), I’ll do my best to build it. Drop your ideas in the comments

Anthropic Opus 4.6 will change the AI game moving forward. Agents don’t fail because they “can’t think.” They fail because they don’t have enough working memory to keep the full problem in view. Opus 4.6 pushes that constraint forward in a very practical way: - 1M context window → you can hand an agent entire repos, months of logs, big contract bundles, or a full research dump in one run. - Long-context retrieval actually works (not perfect, but meaningful): Anthropic reports ~76% on a tough 1M “needles in a haystack” style benchmark (MRCR v2, 8 needles). - Huge outputs (up to 128k tokens) → agents can produce complete deliverables (multi-file plans, long-form specs, audits) without constantly “continue”-ing. Why this matters for builders: ✅ RAG gets simpler (less glue code, fewer brittle chunking hacks) ✅ Multi-agent systems get stronger (the “manager agent” can keep the full state + delegate cleanly) ✅ Coding agents become more reliable (more of the codebase in view → fewer wrong assumptions) ✅ Research agents become usable (more sources in-context → fewer missing citations / gaps) We’re moving from “agents that do tasks” → agents that own projects.

@kanavtwt @zeroblocks If it's trained to do it at the llm level, then it can do it easily, it's not special. We need to see it performing tasks it was not trained on.

> Opus 4.6 wrote a C compiler from scratch which compiled the Linux kernel successfully and you think it's a bubble???