Oslo

Superintelligence is about to emerge.

Introducing 𝑨𝒕𝒕𝒆𝒏𝒕𝒊𝒐𝒏 𝑹𝒆𝒔𝒊𝒅𝒖𝒂𝒍𝒔: Rethinking depth-wise aggregation. Residual connections have long relied on fixed, uniform accumulation. Inspired by the duality of time and depth, we introduce Attention Residuals, replacing standard depth-wise recurrence with learned, input-dependent attention over preceding layers. 🔹 Enables networks to selectively retrieve past representations, naturally mitigating dilution and hidden-state growth. 🔹 Introduces Block AttnRes, partitioning layers into compressed blocks to make cross-layer attention practical at scale. 🔹 Serves as an efficient drop-in replacement, demonstrating a 1.25x compute advantage with negligible (<2%) inference latency overhead. 🔹 Validated on the Kimi Linear architecture (48B total, 3B activated parameters), delivering consistent downstream performance gains. 🔗Full report: github.com/MoonshotAI/Att…

Mercury 2 is live 🚀🚀 The world’s first reasoning diffusion LLM, delivering 5x faster performance than leading speed-optimized LLMs. Watching the team turn years of research into a real product never gets old, and I’m incredibly proud of what we’ve built. We’re just getting started on what diffusion can do for language.

@heyitszeth @shiftj I think this should be the ideal path to test your product out and let a few people use it so that you can pivot or upgrade for the better. Doing the same.

the uncomfortable part is watching other founders seem to move at light speed, they're automating everything with AI, running influencer campaigns, throwing paid ads at it from day 1. Meanwhile I'm in the DMs and manually onboarding one seller at a time wondering if I'm falling behind.

Remember: Do things that don't scale. Automation is great but not on day 1 - just like hiring, don’t automate too early. For example: 1/ Don’t automate your first outbound, do it manually first. You have to feel the responses, see what clicks, and gauge metrics closely. Try to automate that and you won't get the pulse you need to right a truly great outbound. You'll also lose touch with what messaging resonates. 2/ Don’t GPT your customer’s problems, get out and talk to them. It's easy to google search for problems, scrape reddit, or ask Openclaw to do it for you - but there's no substitute for talking to customers. You have to see their reactions and emotions in real-time. That's how you know if you're onto something. 3/ Don’t hand your support to a bot, fix it yourself. Same reasons as above. You need that visceral pain, that tight relationship. These are painful in the short-term, priceless in the long-term. -- Lesson: 1. Do the job until it hurts. 2. Then automate. 3. Then hire. That's how you build a 20x company. Skip steps? you’re left with a clunky setup full of slop.

@3xHarrison but still people will use anthropic or chatgpt. Having a good domain is not related to better research. Meta gave $1 billion to hire individual researchers. Even they don't care about ai.com . It is all hype. Hype creates better visibility but not research.

when you invest in a great domain, you get it when you don't, you don't

@AtomAppraise appraise 8th.ai

@DCLjasonx "why their not" 😆

Got an email from FBI, should i respond?

@PeterDiamandis they design words

One of the most ironic things is that “entrepreneur“ is a French word.

@arpit_bhayani Ask : "are you sure"

I wish my friends were this supportive 🤦

@SumnerLN @mikhailshapiro @sandroherbig @alexblania @sama Congrats 👏

Ecstatic to announce Merge Labs! 🧠🚀 I’m honored to work alongside my co-founders—Tyson Aflalo, @mikhailshapiro, @sandroherbig, @alexblania, and @sama on bridging biological and artificial intelligence. They’re *the* team to take on a mission this ambitious. We’re grateful to have partners like @OpenAI who share a long-term view on what it takes to build foundational technology. If you want to work on hard, foundational neurotech problems, come build with us.

@Sullys_Blog i find : the better my research, the better the people I happen to interact with.

The best drop catching happens before the drop: research and preparation.

@RickdotX congrats to both.

I leased bet.nft💰@ 12 months for $2000. Amazing domain. Congrats to the buyer! #Domains

@SOLsesame @tapdotfun sweet

if you were farming yaps yesterday and woke up unemployed today i'm going to do something stupid i'm building @tapdotfun in public and rewarding early supporters $200 every single day to random interactions on my posts no API for X to revoke indians and nigerians welcome

@katerleonid moves from "i want this" to "let's see who can beat me" - a classic primal urge.

The last 5 minutes of an expired domain auction tell you more about the bidders than the domain itself.

@katerleonid appraised low i suppose. Estibot appraised 8th.ai for $51,000 and i am not sure the exact params it used

Someone bought musical.ai for $464 at an expired auction in 2021. Today it’s appraised at $16,000.

@Ronald_vanLoon @nvidiaomniverse That is THE FUTURE

Physical #AI and #Robotics will bring about the next industrial revolution by @nvidiaomniverse #Innovation #EmergingTech #Tech #Technology

@DCLjasonx Smart move

Owning the exact-match .com is priceless In this example, I removed my numero uno competitor from the equation, because Google now recognizes them as a fake website wordleespañol.com 😂

@brokertom startups usually buy .com after they become successful.

A premium .COM doesn’t guarantee success. But it removes friction, builds trust faster, and gives every other part of your strategy a better chance to work. It’s not the whole engine — it’s the frame it sits in. 🚗

@ben_m_somers I am super on this, also building it solo.

We are building the AI education product that will replace School Looking for tutors & data labelers to help us build it. Rates from $50/hr -> $200/hr. Come help us delete school

@willccbb on point.

nobody has heard of Claude except for the people that spend money on AI products

@SOLsesame @tapdotfun from now on "build in public" = #10F420

times square went green for @tapdotfun