Fluixo

@sultanmeghji yeah concentration never really changed much fr

Five banks printed records today and the press called it “Super Tuesday.” The other ~4,256 FDIC-insured banks weren’t invited. “The banking sector is strong” now means six phone numbers in Manhattan are strong — the same concentration we spent the 80s pretending mainframes didn’t have.

@zayn4pf fr nice one sounds like progress

I’ve been early to a shit ton of good things all my life in crypto, and taken them for granted cos i lacked wisdom. I’m glad i learnt a lot of lessons and built my tunnel vision before i found the best thing, Monad. slowly, then all at once.

@Lummox_eth always good to double check safety with agents

Erik Meijer : "Agents are dangerous until proven safe. You should never let your agents do something unless you can absolutely prove that it's safe." In this 21-minute talk, Meijer shows how proof-carrying code makes agent actions machine-checkable before execution. intent → typed program → safety proof → small checker → execution Formal proofs, not vibes, before the first side effect. Watch it, then read how to build verifiable agent harnesses in the article below.

@vanvster they played great but the final is a different beast

Spain are heading to the final after a 2–0 win They controlled the game, punished France in the key moments France had the names but still Spain looked like the better team A fully deserved result! Are Spain now the favourites to win it all? What's your prediction?

@Ebubededraw nice green vibes on this bird

6m coded

@what3verman kinda funny seeing that bath setup fr

It’s a Spain Yamal thing. They dictated the dictator. 😭

@bitfloorsghost yeah that must feel good

it's wild i'm actually scrolling my own for you feed and organically liking all these posts i see because they are funny and from my friends

@Lummox_eth knockout football is decided by moments not stats

France vs Spain is easily the biggest game of the World Cup so far and I’m backing France they’ve looked consistent from the opening match 2-0 vs Canada 3-1 vs Japan 1-0 vs Morocco 2-1 vs Germany 2-0 vs England 5 matches. 10 goals scored. 2 conceded. they’ve found ways to win every type of game Spain have been excellent as well their possession football has controlled most matches, and they rarely allow opponents to create clear chances with players capable of changing the game in one moment, they’re never an easy team to face but I still give the edge to France they’ve looked more dangerous in transition, defend well under pressure, and have the experience to handle knockout matches this feels like a game that could be decided by one or two moments a place in the final is on the line I’m taking France to win my score prediction is 2-1 and I’m betting on @bananagun

@priyaljain817 check DM pls

narrative precedes distribution always.. without exception

@DaFameless @LunardollzNft @CyberKongz that brick kong dome looks clean fr

Huge shoutout to the @LunardollzNft Apes family! Absolutely loving this brick-bodied Kong with the brain dome. The way this community supports @CyberKongz through the solana:9bmvoZZphKMmNxVbQ2q8AwvHRwENAYcYMTSK39jaCHDa and havihng Henry into their Spaces is incredible. The bridge between Sol and Eth is stronger than ever. Let's keep building! !ooh 🦍⚡🌕

@EugBass up to you

@fluixoo wtf? you mean don't buy mac mini?

Everyone says “buy more VRAM” for local LLMs. VRAM decides what fits. Bandwidth decides whether the model actually feels fast. - VRAM loads the model - Bandwidth feeds every token - Quantization moves less data - Slow memory turns expensive hardware into a waiting room Do not buy an “AI PC” because the memory number looks huge. Check the bandwidth before you check out.

@beamnxw small hardware is getting surprisingly capable

WATCH THIS DEVELOPER SNAP A GRAPHICS CARD INTO A POCKET-SIZED SINGLE BOARD SERVER TO RUN LOCAL LLMS If you think running local models means you have to buy a massive tower case or an overpriced server rack, you need to watch this clean desktop setup. This guy just unboxed the new ZimaBoard 2, a pocket-sized single-board server that completely challenges the traditional form factor layout The hardware engineering on this little silver brick is wild. It features a sleek, ribbed metal chassis that acts as a giant passive heatsink to keep things completely quiet. It comes packed with standard connectivity inputs, dual high-speed LAN ports, and an external interface slot right on the side panel The coolest part is watching him unbox a low-profile Gigabyte graphics card and snap it directly into the server board expansion port. The cardboard packaging even folds out into a clever, minimal desktop stand to house the entire combined rig securely. Once he plugs in the power and display cables, the custom cooling fan kicks on instantly, transforming this tiny footprint into an off-grid processing station It is a remarkably clean, low-cost approach for developers trying to host dedicated local networks or build portable testing nodes. Why build a massive, noisy desktop rig when you can assemble a fully functional local node right on your coffee table

@Di_Krass_ cheaper compute changed everything

Jensen Huang - CEO of NVIDIA, at Stanford. for 64 years computers worked almost the same. then NVIDIA sped up computing a million times in 10 years - and that's what made all of AI possible. computing used to be "pre-recorded" - now it's generated in real time. machines learned not to draw, but to "think." the point isn't faster chips. when computing gets a million times cheaper, engineers stop curating data and just feed the AI the whole internet. it's not the old world anymore - it's a new one. 68 minutes with the man whose chips power the entire AI industry. Follow me to know more. Save the post so you don't lose.

@karpachoq thats some specific prompt engineering right there

Boris Cherny would recognize the 9:17 moment when 6 cleanup tickets stopped moving to next week and became 1 clean Claude Code run. 40 minutes setup. 1 narrow task. 10-file stop sign. 3 failed-test cap. The engineer did not ask Claude to “help with the project.” He pointed it at src/auth/, named the duplicate session parser, protected the public API, and demanded 1 regression test. That changed the session before code existed. The repo had walls, the agent had edges, and the reviewer had evidence instead of a polished story. Small issues stopped becoming furniture once the agent could only touch the room it was sent into. x.com/karpachoq/stat…

soon beta test!!! @FirsayAI BD, outreach managers, and others involved in research will soon be able to try this out

@zarqXBT that's a game changer for agent reliability

This paper completely changed how I think about agent memory: Current agents forget everything between sessions -> this framework treats memory as durable state -> audit every decision -> expire stale memories before they trigger bad actions -> govern what the agent remembers. That loop is why enterprise agents keep hallucinating from outdated context while this architecture does not. Persistent state + memory governance + auditability + provenance tracking - that's the missing layer. Read and save it, then rethink how your agents handle long-term memory.

@fleyta88 ai is getting wild ngl gotta stay sharp

Google DeepMind CEO, Demis Hassabis: "AGI is now on the horizon, and it will be the most profound and impactful technology ever invented. When we look back at this time, I think we will realize that we were standing in the foothills of the singularity." In this Google I/O 2026 keynote, Hassabis unveils Gemini 3.5, AlphaEarth Foundations, and Gemini for Science, aimed at one day solving all disease. Watch it today, then read how AI simulations are already predicting hurricanes before they form in the article below.

@tsukiema_ wow 1961 that's wild to think about

The Thinking Machine (Artificial Intelligence in the 1961): "Can machines really think?" This short film from 1961 is one of the first public discussions of AI. Straight interviews at MIT You see a computer playing checkers, translating Russian to English, and Claude Shannon and Oliver Selfridge explaining what intelligence would look like in a machine They predicted within 10 to 15 years machines would handle tasks once reserved for human intelligence Bookmark this. Its 3 minutes and worth more than most new docs

@gippp69 this could be huge for dev teams lowkey excited

SOMEONE GAVE CLAUDE FABLE 5 A CONTRACT, A $5 DAILY BUDGET, AND A MANAGER. THIS F**KING DANGEROUS SYSTEM CAN CLEAR A WEEK OF BACKLOG WHILE YOU SLEEP. 00:03 he opens Microsoft’s 4,900 star Agent Governance Toolkit, built for policy enforcement, isolated execution, identity controls, and protection against all 10 OWASP agentic risks. setup starts with 3 files. CONTRACT MD sets the limits, boundaries MD defines what the agent can touch, and signoff SH runs every test before anything ships. four roles split the shift. one model reads the logs, Fable picks the highest value task, another writes the code, and a fresh Fable reviews the final diff. a $5 cap stops runaway sessions, commits above 150 lines need approval, and autonomy unlocks only after 20 runs at a 95% pass rate. one failed run sends the agent back to probation. the repo adds the guardrails, but the contract is what turns Claude from a chatbot into an employee.

@diamai_ solana ai

Sol won the Milky Way and lost the apartment. Pat Simmons ran ten side-by-side tasks across Sol, GPT-5.5, Opus 4.8, and Fable 5. The lead changed with the work. 01:18 He separates the GPT-5.6 family into three jobs. Sol handles ambitious agent work. Terra sits in the middle at roughly half the price of GPT-5.5. Luna is for quick recurring work such as call summaries and email triage. Model choice is only one setting. Max gives one model more time to work. Ultra starts four agents in parallel by default. 16:48 Sol wins the interactive Milky Way map after its first Rubik's Cube build failed to render and its apartment reconstruction put the rooms in the wrong places. The apartment looked polished until you checked the layout. 38:07 Fable finishes slightly ahead on the coding builds. Sol wins the knowledge-work section, including the PowerPoint and airline rebrand. Pat ends up recommending a hybrid setup where different models handle different parts of the build. The article below turns that uneven scoreboard into a working router. It includes a copyable task card, five model roles, proof rules, retry limits, permission ceilings, and a 20-run calibration loop for deciding when a job moves up or down. Keep the task and its checks stable. Change one layer for the next five runs.