Adam Killam

Judging by my tl there is a growing gap in understanding of AI capability. The first issue I think is around recency and tier of use. I think a lot of people tried the free tier of ChatGPT somewhere last year and allowed it to inform their views on AI a little too much. This is a group of reactions laughing at various quirks of the models, hallucinations, etc. Yes I also saw the viral videos of OpenAI's Advanced Voice mode fumbling simple queries like "should I drive or walk to the carwash". The thing is that these free and old/deprecated models don't reflect the capability in the latest round of state of the art agentic models of this year, especially OpenAI Codex and Claude Code. But that brings me to the second issue. Even if people paid $200/month to use the state of the art models, a lot of the capabilities are relatively "peaky" in highly technical areas. Typical queries around search, writing, advice, etc. are *not* the domain that has made the most noticeable and dramatic strides in capability. Partly, this is due to the technical details of reinforcement learning and its use of verifiable rewards. But partly, it's also because these use cases are not sufficiently prioritized by the companies in their hillclimbing because they don't lead to as much $$$ value. The goldmines are elsewhere, and the focus comes along. So that brings me to the second group of people, who *both* 1) pay for and use the state of the art frontier agentic models (OpenAI Codex / Claude Code) and 2) do so professionally in technical domains like programming, math and research. This group of people is subject to the highest amount of "AI Psychosis" because the recent improvements in these domains as of this year have been nothing short of staggering. When you hand a computer terminal to one of these models, you can now watch them melt programming problems that you'd normally expect to take days/weeks of work. It's this second group of people that assigns a much greater gravity to the capabilities, their slope, and various cyber-related repercussions. TLDR the people in these two groups are speaking past each other. It really is simultaneously the case that OpenAI's free and I think slightly orphaned (?) "Advanced Voice Mode" will fumble the dumbest questions in your Instagram's reels and *at the same time*, OpenAI's highest-tier and paid Codex model will go off for 1 hour to coherently restructure an entire code base, or find and exploit vulnerabilities in computer systems. This part really works and has made dramatic strides because 2 properties: 1) these domains offer explicit reward functions that are verifiable meaning they are easily amenable to reinforcement learning training (e.g. unit tests passed yes or no, in contrast to writing, which is much harder to explicitly judge), but also 2) they are a lot more valuable in b2b settings, meaning that the biggest fraction of the team is focused on improving them. So here we are.

We're bringing the advisor strategy to the Claude Platform. Pair Opus as an advisor with Sonnet or Haiku as an executor, and get near Opus-level intelligence in your agents at a fraction of the cost.

wen spud

@trq212 Interested!

I want to do some streams where I work with non-technical people using Claude Code to figure out how they might be able to improve their process. My feeling is that just a few tips could make a big difference in efficiency. Any mutuals interested?

Go ahead and steal the playbook, it’s free

@sophianabilg When is Vancouver registration opening?

Friends - I’m going on a little Lovable tour!!! Today we’re launching the Lovable Founder Series, coming to 50+ cities worldwide. The Founder Series is all about showcasing what founders are creating with @Lovable and connecting them with each other. In every city, we’re bringing together builders, founders, and creators to show how you can actually start and scale a business with Lovable. Over April and May, I’ll be traveling across some of our main hubs to meet our builders. Having met so many of you online, it feels really special to now connect face-to-face. We’re kicking things off in Stockholm today, and heading to Paris tomorrow to continue the tour. Here are a few of the stops: 🇸🇪 Stockholm, April 7 🇫🇷 Paris, April 8 🇩🇪 Berlin, April 11 🇪🇸 Barcelona, April 15 🇺🇸 Boston, April 17 🇪🇸 Madrid, April 18 🇬🇧 London, April 21 🇮🇹 Milan, April 23 🇵🇹 Lisbon, April 25 🇳🇴 Oslo, April 27 🇩🇰 Copenhagen, April 28 🇫🇮 Helsinki, April 29 🇳🇱 Amsterdam, May 4 🇺🇸 NYC, May 6 🇺🇸 LA, May 8 🇺🇸 SF, May 12 And this is just the beginning…..we have 50+ events happening globally, powered by our amazing community! If you’re in any of these cities - come join us. Would love to meet you <3 See you on the road ;) Link in the comments to find your city

This is genius level comms… I wonder if they used AI 😆

@alexandr_wang Where can we try it out?

1/ today we're releasing muse spark, the first model from MSL. nine months ago we rebuilt our ai stack from scratch. new infrastructure, new architecture, new data pipelines. muse spark is the result of that work, and now it powers meta ai. 🧵

I know what I’m doing this weekend

@audrlo Sam

I just built a device to keep my grandma safe. Introducing Sam, the first AI caretaker for seniors. It talks to your senior, monitors their cognitive health, and lets you know if something's wrong. If you want one for your grandma... Comment "SAM" and we'll send you one.

Never bet against him

@yacineMTB Canadians had a chance to vote for change but 8.5 million decided they wanted more of this: more decline, more government, more virtue signalling 🤷🏻‍♂️

canadians shouldn't be this poor given the natural resources we have available

@aakashgupta I thought this take was going wildly off base until the end. Had me in the first half, not gonna lie

The math on AI agents and the web is so lopsided it should embarrass every "API-first" pitch deck in Silicon Valley. 1.1 billion websites exist. Roughly 50,000 have a public API. That's a coverage rate of 0.005%. Every AI agent demo you've ever seen runs against the 0.005%. Booking a flight on an airline's API. Pulling data from Salesforce. Sending a Slack message. Clean, structured, predictable. Then you ask it to renew your driver's license, check inventory at a regional distributor, submit an insurance claim, or pull data from your kid's school portal. The agent hits a login page rendered in 2009-era HTML and dies. This is the gap Browserbase is building into. $67.5M raised, $300M valuation, 50 million browser sessions, 1,000+ customers. The bet: AI doesn't get a clean internet. AI gets THIS internet, with CAPTCHAs, cookie banners, JavaScript rendering, bot detection, and session management. The SaaS industry spent 15 years building walled gardens. AI agents need to climb the walls. The browser is the ladder. Browserbase is building the Twilio for headless browsers. And if you understand what Twilio did to telephony infrastructure, you understand why Kleiner Perkins and Patrick Collison are writing checks.

Introducing Project Glasswing: an urgent initiative to help secure the world’s most critical software. It’s powered by our newest frontier model, Claude Mythos Preview, which can find software vulnerabilities better than all but the most skilled humans. anthropic.com/glasswing

@ExaAILabs @coinbase What will it cost per search?

We're excited to partner with @coinbase to enable agents to natively pay for web search, via x402! x402 is an open protocol that enables agents to pay via HTTP, governed by the Linux Foundation. When an Exa API request is made without an API key, Exa now returns a 402 status code with payment information that an agent can act on.

@hnshah I thought this was a post about Canada for a second 😆

Your best people are leaving. If it was just money, you could fix it. This is different. The real reason is your environment doesn’t let them do their best work. You hired A-players and they became spectators. You hired people with taste and speed, then trained them to wait. And instead of fixing the environment, you added more process. More oversight. More “alignment.” The result? Slow suffocation disguised as management. I've been on both sides of this. Built environments as a founder. Lived in one as an IC at Dropbox. The difference taught me everything about why great talent fails in wrong environments. Great people don’t suddenly get mediocre. The environment makes them act that way. I wrote about the real framework that works: Vibe → Environment → Culture. How @morganb made his explicit. And why this order changes everything. If you're losing people you can't afford to lose, this might explain why.

Magical OpenClaw experiences that use frontier models cost $300-1,000/day today, heading to $10,000/day and more. The future shape of the entire technology industry will be how to drive that to $20/month.

Is selling agents that build businesses the equivalent of selling online marketing courses?

@IMJustinBrooke Interested for sure

Before they change the definition on us… Here’s what ANI vs AGI vs ASI means (or meant) originally. ANI is what we are currently experiencing, we just call it AI though. It stands for artificial narrow intelligence and means it’s AI that’s really good at specific tasks. Kind of like how Claude Code is really really REALLY good at coding. Or Claude Sonnet being great at writing. This is why there are soooooo many AI tools. Each has to be specifically trained (fine tuned) on a specific task. Then there is AGI. This is the one all the big tech companies are racing towards. It’s the reason HUNDREDS of billions of dollars have been invested. AGI stands for artificial general intelligence. Meaning AI will be better than humans at most/all “general” tasks. This is the “moving” definition. I say moving because 4yrs ago when I first got into AI, it meant all general tasks. Today it’s been bent a little to mean “most” tasks. And I believe in a few months Anthropic (maybe another company) is going to claim AGI but the definition will be bent even more. My needle points to Anthropic specifically because they invented “agent skills.” Millions of people are uploading their S.O.P.s, best practices, proprietary methods, and recipes into Claude via their skills. Now they have plugins which are like advanced skills that do a wide variety of things. Soon (next two big updates) they’ll probably have categories or roles. It’ll be an update that says their AI can do all the skills in whole roles or categories. Like marketing, or finance, or health etc. We are super close to this. It could be a month away. It could be tomorrow. But I’m thinking Fall season is going to be massive. These companies will want to have record numbers for the year. They’ll work hard all summer for Fall launches to give enough room for news cycles and 2mos of growth. Which leaves us with ASI. This is artificial super intelligence or what y’all would refer to as SkyNet. That’s the one that will be better than humans at pretty much everything, but not marginally better… exponentially better. So much better that our minds can’t even grasp at how much better because we don’t yet know what we don’t know. For example, some people can taste colors. It’s a real disorder or phenomenon (forgive my ignorance, not a doctor). What does yellow taste like? What does light blue taste like? Your mind has no context for it, so you can’t even imagine. ASI will literally be off the charts. It’ll be like a copywriter who can predict in real time the exact words to use for every individual visitor to your website because it just read every social media post they’ve ever made. (Someone build that!) This will require quantum computing. The tech of today is just not physically possible to get there. Really we can’t get to AGI either, but the definition is covered in more baby oil than P. Diddy has in his basement. It’s slippery and getting more slippery by the week. I won’t be surprised if someone from big tech tries to beat everyone with a blow out summer announcement. But just remember ChatGPT was announced in November. So was Kimi K2, Gemini 3, and Apple AI. It’s not accidental. Lastly, one more prediction. Next year will be the year of local LLMs. The models are good enough now to compete with the big subscriptions. They’ve also become efficient enough to run on common machines as well as computer companies developing AI ready consumer computers. The big 3 will raise prices, it’ll bring local models mainstream instead of just extremists like Finn and his cult following. They have to raise prices, it’s getting harder to scale on hardware alone. I’m hoping for a $1,000/mo mega max plan. I’ll be first in line. Already tapping out my $200/mo plans on Manus, Gemini, and Claude. Launching “Build Notes” this week. It’ll have news like this plus detailed AI recipes. Interested? Let me know…