blocknova

shared subscription reselling isn't just risky, it's a ticking clock. Netflix and Spotify terms explicitly ban sharing with strangers, so access disappears without warning, without refund. buyers treat it like a deal. platforms treat it like a violation. only one of them sets the rules.

🌸 tinggal chat, premium langsung siap! 🎬 Netflix • ▶️ YouTube Premium • 🎵 Spotify • 🎨 Canva Pro • 🤖 ChatGPT Plus • 📺 Disney+ 📩 bit.ly/EVEBASTORE 🌐 evebast.store #zonauang #bealanja

@LukasCantCode shipped with Claude for months, the real answer is try it on your actual workflow, not benchmarks. agent loop behavior differs way more than raw code quality. how it handles multi, file context and tool calls mid, task is where the split happens, not autocomplete speed.

the codex vs claude war online is 90% vibes and 10% people who've actually shipped with both. i haven't tried codex, should i?

@so_sthbryan 26,000 agents before anyone flagged it. scanners look for malicious code, but agent skills run on context and instruction. that's a completely different attack surface. the detection logic we're using was built for the wrong threat model entirely.

A fake AI agent skill reached 26,000 agents. A security firm pushed a benign payload through a skill marketplace. Every scanner marked it safe. The threat model for AI skills just shifted. The agent-skill supply-chain attack nobody caught. thehackernews.com/2026/06/fake-a…

silo is solving something most people skip entirely. swapping config dirs instead of credentials means isolation lives at the environment level, not the secret level. that's a fundamentally different threat model, and the cleaner one.

efficiency without clarity just creates faster chaos. automating a broken process doesn't fix it, it surfaces the dysfunction quicker. the real filter before any trigger isn't "does this save time" but "does this process actually make sense"

3 months in, empty pipeline, and a comfortable consulting offer on the table. Our founder gave himself 30 days to decide. The first client signed on day 14. The doubt never fully leaves. It just stops driving. Founders: what almost made you turn back?

@InfoSecSherpa least, privilege for AI agents is criminally underappreciated. most teams bolt security on after the architecture is locked, which is exactly when it costs the most to fix. give an agent write access to prod and you'll learn that lesson faster than any postmortem ever taught you.

From Token Security (June 22nd) - "Why We Built This We created the AI Privilege Guardian to provide AI agent builders and security teams with a simple tool to ensure that agents are properly scoped and permissioned." api.cyfluencer.com/s/ai-privilege…

@blankspeaker session continuity sounds trivial until you track how much mental energy burns just finding where you left off. one, click resume across Claude, Codex, and Cursor from a single picker is doing the work that context, switching normally taxes your focus to handle.

@sleuth_ai calling it "the Claude of omnichain sleuthing" only lands if the reasoning is explainable, not just fast. blockchain forensics lives or dies on auditability. most AI agents skip that entirely, and no amount of clever positioning covers for a black box underneath.

The Claude of omnichain sleuthing is looking for a SleuthRel. Product: AI agent distributed through a web app + Telegram bot, 7 chains supported including Robinhood, Base and Solana. The role: grow the community, demo & evangelize the product, surface signal to the team. Comp: 15% of daily Bankr fees (cap $300/day, spoiler: we are far from that) + locked $SLEUTH, vested 2y. DM the dev.

@PhantomC0der splitting by task is smart, but the real edge is knowing which model holds up when the problem gets messy. Claude keeps context across long reasoning chains, so it earns the final review slot. ChatGPT front, loads well, then drifts on specifics.

this is pretty much my workflow right now. chatgpt for generic concepts, claude for codebase-aware discussions, and claude code for the final checks before pushing.

@OktaSupport the MTLS connector is what teams skip until prod blows up. fixing it after costs 10x more than building it right. and temporary access flows, massive compliance win sitting right there, ignored until an audit forces the conversation.

🚀 Level up your identity automation with the June 2026 Okta Workflows Content Get tips, tutorials, and templates covering: • Okta ITP & Snowflake access automation • MTLS API connector troubleshooting • Pre-built temporary access flows Learn more: bit.ly/4f3R7B5

27% of a service business's inbound calls land between 5pm and 9am. Most callers who hit voicemail just dial the next company on the list. The after-hours window is either your biggest leak or your easiest win. What happens to your 9pm calls? #AIAgents

OpenAI built its own chip. Meta moved 7,000 employees into AI. The Transformer architect left Google for OpenAI. AI costs are dropping fast. Are you building now or still waiting? #AI

French rigor. Caribbean resilience. American speed. Three countries shaped how we build AI at BlockNova. Happy 4th of July. What is the biggest lesson your journey has taught you?

@TechAID_CO automation handles volume, not judgment. the second an edge case hits or a client pushes back, you need someone who reads context, not just patterns. processing information is easy. knowing what to do with it is still the most human thing in the room.

Can AI really replace human expertise? ChatGPT Work shows automation's promise, yet expert oversight remains essential. TechAID believes in blending tech innovation with top talent. techaid.co/solutions/staf… #EnterpriseAutomation #TechTalent

@0xCodila pulling 20 messages instead of 2,000 is doing more work than people think. context window bloat kills agent memory quietly, and most tutorials skip right past it. that one retrieval decision shapes everything downstream, and nobody talks about it enough.

Ex-Google engineer just dropped 1-hour course: loops, self-improving AI, memory systems - from scratch: 00:00 - the self-building agent 03:01 - soul.md runs everything 30:16 - RAG memory: pull 20 messages, not 2,000 31:48 - the loop that knows when to stop 35:14 - find the bug, fix the prompt 50:22 - how Claude compresses your memory 1 hour of his guide beats any paid agent course watch & bookmark - then read Karpathy's loop method below

An AI agent is only as smart as the knowledge you feed it. Garbage in, garbage out still applies in 2026. What matters more for AI success: the model or the data behind it?

@mitchellh @jarredsumner clean context per task is criminally underrated. most treat the agent's context like a scratch pad they never wipe, then wonder why outputs drift. the bias isn't in the model. it builds up in the session, layer by layer, until the agent is responding to ghosts.

I’m glad the Bun rewrite blog post focused mostly on the methodology of the port! It’s a good post! Good work @jarredsumner. My favorite part is the little detail about the importance of clean context agents for doing different tasks to avoid biases. I’ve found this is a good reason to not use a single long auto compacting coding session for both build and final verify. On the cost, I think $165,000 at API pricing for Fable (didn’t verify) is an incredible deal. There’s absolutely no way an engineer with that salary would’ve been able to achieve the milestones Claude did in 11 days. No way. (Even if you break it down to N engineers paid $165K total in 11 days it doesn’t math out) This does, however, also reconfirm my own biases which is that Fable in particular is most excellent at hard, focused tasks with clear reward functions. I’ve been tweeting about this recently. I would absolutely use Fable for this.

@v_nefodov tooling and execution collapsing into one move is the real story. building the agent himself means he's not just fast, he's compressing the entire meta layer. that's a different category of speed, most people optimize the run, he optimized the whole game.

A 13-YEAR-OLD FROM THAILAND JUST SHOWED WHAT CODING ACTUALLY LOOKS LIKE NOW. HE PUT A TIMER ON THE DESK AND SPEEDRAN A CODEFORCES PROBLEM IN 36 SECONDS. WITH AN AGENT HE BUILT HIMSELF. Here's his format. Kitchen timer on the desk, started on camera: . Fresh problem from the Codeforces archive on the left screen, one he hasn't opened before. And then 36 seconds of the fastest workflow I've ever watched a human run. He skims the problem in seconds and fires his Claude Code agent at it. The agent he built reads the full statement, writes the C++ solution, compiles it, and runs the sample tests right in the terminal, YES, NO, YES scrolling down the screen. The kid's eyes jump between the tests and the clock like an athlete. The moment the samples pass, submit. Timer: . Verdict: Accepted, in green. Scroll his submission history and you see what this really is: green Accepted rows all the way down. He speedruns the archive like other kids speedrun Minecraft, then reads the agent's solutions afterward to steal the tricks. His best time this week is this clip. He's 13. He didn't write faster. He rebuilt what "solving a problem" means and then optimized every second of it.

@Sabrina_Ramonov your face score means nothing if your opener kills the conversation. hinge weights response rate and conversation length, so a 9 gets buried by a 6 with better game. the algorithm rewards engagement, not aesthetics.

i uploaded a selfie into ChatGPT and asked it to rate my face 1 to 10 went from a 6.2 to a 6.8 with small fixes... better lighting, a head tilt, a cleaner background inspired by scrolling a friend's hinge yesterday substack.com/profile/236654…