Matt Kunihiro

Has stackoverflow but for agentic work vibes

@JoshKale Didn’t they just accidentally leak their own source code map two weeks ago? Are they using mythos on themselves?

This is big... Anthropic just announced a model so powerful they won't release it to the public out of fear over the damage it will cause 😨 Claude Mythos Preview found thousands of zero-day exploits in every major operating system and web browser... The numbers are hard to believe: > $50 to find a 27-year-old bug in OpenBSD, one of the most security-hardened operating systems ever built > Under $1,000 to find AND build a fully working remote code execution exploit on FreeBSD that grants unauthenticated root access from anywhere on the internet > Under $2,000 to chain together multiple Linux kernel vulnerabilities into a complete privilege escalation exploit For context: these are the kinds of findings that previously required elite security researchers working for weeks. Anthropic engineers with no formal security training asked Mythos to find exploits overnight. They woke up to working code the next morning. The results were so impressive Anthropic assembled Apple, Google, Microsoft, Amazon, NVIDIA, and seven other organizations into Project Glasswing: A $100M defensive coalition. They're not releasing this model publicly. Instead, they're racing to patch the world's infrastructure before models like this proliferate.

@samhenrigold We must shame this into oblivion haha

i know what you are.

@samhenrigold @NAndriievskyi thank you!

@NAndriievskyi it looks like poop from a butt

This is a Music Typewriter from the 1950s.

Beautiful animations to start. Take some time to celebrate this milestone. Some ideas when you're ready. Would be awesome to be able to port in from some of the other moodboard platforms if possible/accessible APIs (pinterest, myMind, Milanote). I know that might not be so easy though.

@tbsrnstrm congrats, man. been watching from the sidelines for a while as you share updates. it looks beautiful. excited to try it out

Almost 16 months in the making, I'm finally at a point where I'm calling this good to go on the App Store. It's been my privilege to share the journey with you all, here's to the cool things still to come. Spatial is now available at apps.apple.com/app/id67539798…

My dear front-end developers (and anyone who’s interested in the future of interfaces): I have crawled through depths of hell to bring you, for the foreseeable years, one of the more important foundational pieces of UI engineering (if not in implementation then certainly at least in concept): Fast, accurate and comprehensive userland text measurement algorithm in pure TypeScript, usable for laying out entire web pages without CSS, bypassing DOM measurements and reflow

@benjitaylor I cannot wait you are the right guy for the job

First thing on the agenda: improve everything

@scottbelsky Reddit is considering this already

how long until we’ll need to give an app or website keyboard access and use FaceID for “proof of human” commenting, purchasing, and form completion?

A lot of PMs that built their careers on documentation and rituals are going to struggle if they don’t embrace the fact every role’s artifacts are moving toward the source of truth itself: the code.

@danywander @georgiNikoloff 🙌🙏 thanks so much !

@mattkuni @georgiNikoloff svg > plasticity > zBrush > blender > substance painter + photoshop > blender > glb :)

claude code is amazing. literally one shotted this badge preview to test glb models. kudos to @georgiNikoloff for teaching me tree js stuff

@igarakii 🔥

motion design reel from last year #motiondesign #aftereffects

I agree. It’s sort of in the same ballpark of moonshot scenarios that are only possible if a ton of things happen that are incredibly unlikely to happen. Time constrains all things, and people will never have enough time to build things themselves, as things worth using will always take time.

My prediction is that “malleable software” has the same fate as low code tools. Artifact of the past

I sort of see two under thought future paths for software. The “SaaSpocolypse” narrative is clear: subscription software multiples collapse in a world of existential competition from the “do every job ever” AI specter. What it has triggered is a wave of PE and VC freak outs, that are jamming AI onto every roadmap and every employees hands in their portfolio. They’ve seen what it’s doing to coding and tech and they’ve extrapolated that out to all knowledge work. They’ve chugged the Amodei Kool Aid. But what if it turns out that AI works really well for individual contributors and tech organizations, but other knowledge work that doesn’t live in an ecosystem of code/libraries/only has a pretty sticky local maximum of accretive value to AI adoption? What does knowledge work productivity improvement really look like past what Claude Cowork is today? More MCPs? More connectors? More skills? Better harnesses? Have the foundational rails already been laid and is there much more to disrupt? In every company touched by AI adoption I’m seeing people work more not less. But I’m not seeing people use less software. I also ask myself, what if AI user experience doesn’t really have that many variants in practice? What if every company jamming it into their roadmap doesn’t really need an agent? What if most AI value belongs at the data level? Or the integration layer between systems? What if we find out that new AI feature popping up on all of our apps that looks the same isn’t that useful? And then I ask myself what happens if AI doesn’t do every job well? What happens if now we just have AI usage-based pricing and our apps sort of suck because of it? What if AI doesn’t primarily replace jobs done by humans but instead by crowding out budget as token OpEx skyrockets? What if AI is feature and a tool but not a transformation? That’s if the promise plateaus. Then think about if the Saaspocolypse does hit. What then? If SaaS valuation multiples stay suppressed and AI continues to prove it’s competitive might. If SaaS companies begin to lose revenue and market share to the model companies and AI first competitors. What will happen in the industry? PE is looking at big draw downs already. Some of the biggest names have seen double digit losses this year. They may look to offload assets, and we may see an increase in m&a transactions. There may be a future strategy emerging to merge as many like-companies into one. Build SaaS monopolies within each niche industry and then layer over MCP/CLI usage based access. Unify and charge fees. What if the industry gets more concentrated not less?

Agreed. The lock in for collaboration, larger org procurement, teams, files is big. It’s all there. I still haven’t seen a better canvas tool. Even though Paper’s doing cool things. I think the risk they run by not moving faster will be letting the canvas itself become a secondary surface in the product design stack rather than the primary one. On our team we see about half/half at the moment. Half of our clients are moving to code first with figma as a support tool, and half are still fully in figma.

i have a feeling most people who are experimenting with other tools will come back or at least add figma back once it catches up. there are definitely some designers who are essentially becoming developers now and it makes sense that they may not come back but i think anyone who does canvas-based work is probably going to stick with figma. especially because of the collaborative lock in. if you're just making stuff for yourself or a team you control you can choose your stack but it's a lot harder the more stakeholders you get.

i honestly think figma hasn't released a design agent yet because the tech just isn't able to do it with consistent excellence. seeing the results from stitch etc is reinforcing this for me.