Very @beh_zod coded behzod.com/blog/being-hired-…
itsbdell 🔎 retweeted
This guy is a closer. Laser focus on the anti-Norway vote. teapawt.org/clashreport/status/211…
“User comfort” is a great framing.
All UI is becoming onboarding
Front-end is sacrificial
The only point of a UI is user comfort the first and only time they see it. From then on:
- personal agents don’t care about the UI
- security means you need to reduce the attack surface and harden
Headless everything 👇
interconnected.org/home/2026…
itsbdell 🔎 retweeted
The most interesting AI search question is not "how to measure citations" it's:
"Should our brand have an MCP?"
Just today - Walmart announced their MCP while Amazon continues to block Muse (and other agents).
And on the Spotify earnings call this morning "As listening shifts to third-party AI agents, as with the new Muse integration, who owns the customer relationship and the data, and how do you keep that from turning Spotify into a back-end supplier?"
This is a critically important conversation with lots of nuance. I recorded a full podcast here, check it out:
newsletter.brxnd.ai/p/should…
itsbdell 🔎 retweeted
When I first met @LorenBrill she had lofty ambitions. She wanted to 50x revenue and get on @guyraz to share her journey. I was skeptical (to put it mildly!) but I saw in her a passion that was undeniable. Today my wife ticks off another box in her life story. Give this episode of @HowIBuiltThis a listen if you want to hear a story about grit, determination, positivity, and resilience. I’m incredibly proud❤️
open.spotify.com/episode/20v…
itsbdell 🔎 retweeted
if marketing really is 80% of the success, here's where i'd spend the 80%:
1. get painfully clear on positioning. one sentence should tell me WHO this is for, WHAT it helps them do, and WHY i'd choose it over the alternatives.
2. steal your language from customers. sales calls, reviews, support tickets, Reddit, DMs. save the exact words people use for pains, objections + desired outcomes and write from those.
3. pick one distribution channel and get unusually good at it. X, YouTube, LinkedIn, short-form, SEO, whatever already has your buyers. depth beats being mediocre everywhere.
4. spend almost as much time packaging as creating. hook, title, thumbnail, first 3 seconds. nobody gets value from something they never click.
5. build around repeated problems, not random content ideas. if customers ask the same question 20 times, that's probably a post, landing page, ad, email + sales asset.
6. study your outliers. when one post gets 10x the normal reach or one ad crushes everything else, don't “move on.” figure out the topic, hook, format + audience behind it and make more.
7. turn one good idea into 10 pieces. article → X posts → video → clips → newsletter → carousel → sales email. good ideas should travel.
8. collect proof constantly. screenshots, customer quotes, numbers, before/after, use cases. marketing gets WAY easier when you don't have to manufacture credibility.
9. move people somewhere you own. social gets you discovered. email, community + your site let you reach them again without asking an algorithm for permission.
10. keep the loop running:
customer → insight → content → attention → conversation → sale → more customer insight.
making something great is only half the job.
people have to know it exists.
itsbdell 🔎 retweeted
.@OpenAI fell behind in the coding race. But then it came back with Codex.
@danshipper tells the story in “How Codex Happened,” drawing on first-person interviews and his own experience: “I was there for a lot of it.”
Subscribe to receive it in your inbox: every.to/subscribe?utm_sourc…
The future of building in public will be built with an AI documentarian.
Ken Burns-Tokens over here @every
itsbdell 🔎 retweeted
Tired: sending an email update to your boss
Wired: sending a Claude artifact showing your work
Inspired: having Claude interview you to create your own personal documentary
👀👀👀
we're dropping something heavy tomorrow @every
it's the 8k word definitive history of Codex
written from deep interviews with the team that did it including @gdb @thsottiaux @ajambrosino @embirico @mia_glaese @edbayes and more
it'll also be agent-native so you and your agent can explore the full text of all the interviews in addition to reading the story :)
Really interesting to see how @attio is thinking about agent-native architecture for CRM. Congrats to @nicolasosharp and the team on the launch!
In the age of agents, systems of record have never mattered more. They’ve also never been less prepared for what’s coming.
Today, we’re announcing @attio's hybrid architecture, with the controls of application software at the scale of infrastructure.
To agents, Attio looks like a database, queryable in SQL - a language they're RL’d on. Unlike a database, the business logic, policies, and compliance sit behind the query.
This makes all of Attio’s data accessible by agents in the most powerful and flexible syntax. It gives GTM teams an order of magnitude more value from their time, with the full power of agents working across the entire customer lifecycle.
A new shape of software is here.
attio.ai/architecture
itsbdell 🔎 retweeted
in place of “alignment” which is impersonal and abstract id like to propose a new key dimension of model performance:
Fidelity: how well does a model act on your behalf to accomplish your ends and show you the pieces of information relevant to you to make decisions.
A high fidelity model acts as an extra sense organ and limb for YOU on your work etc. A model doesn’t get to be high fidelity if it scores well on SWE bench. Only if it provably works well on YOUR work.
imo a large part of the problem with swarms and our ability to monitor and control them is that models are currently low fidelity for this use case: I can’t be confident they would notice what I would notice if I looked at a swarm trace, and I can’t be confident they would do what I would do.
rather than measuring hiw close a mode is to replacing you, measuring models on fidelity measures how well they HELP you do a job better
high fidelity models are the future of models that allow humans to do better work
Dusting off this tweet from April because this gap in shared understanding of LLM capability is *widening*. It's now less "two groups of people speaking past each other" and more a sharp funnel.
- Napkin math somewhere around 6B people (~75% of the population) have barely come in contact with LLMs at all.
- Around 1-2B (~20%) are casual and infrequent users of free-tier ChatGPT-like products. This group sees derpy chatbots and treats them a bit like a better Google search, a writing aid, or etc. Maybe an agent tries to book you a flight. I have non-tech friends who (reasonably, imo) say they have not much use for it at all. Even many professionals outside of math&code are in this tier. For example, execs and many other functions spend a lot of their time talking to other people, so while they understand what is happening intellectually, it is still second-hand and a bit abstract.
- Now we get to professional use of frontier-grade LLMs in math&code. Somewhere around 20M people (0.2%) see first-hand that large, complex projects that used to take them weeks/months can now be completed by agents with a prompt. Building apps, copying apps, translating apps, decompiling apps from binaries... This has all happened very quickly and recently - less than 1 year ago, I was writing code manually by hand, typing memorized computer code commands into a code editor character by character, occasionally pressing Tab to autocomplete a little chunk of code.
- And finally we get to the ~5,000 people (~0.00006%) with access to frontier-grade systems internally. The external world has seen the preview. It looks like swarms of thousands of agents collaborating over weeks on software mega projects: minting zero days, running cyber attacks and defenses at machine speeds, discovering new science, advancing the frontier of mathematics. Things that would have taken top professionals in the industry years of work. Meanwhile, human review and comprehension are starting to fall behind. For example, people are still involved in the "archeology" of the OpenAI-HF incident from many months ago. Mathematicians may be poring over the 722 manuscripts on frontier mathematics for a while.
The funnel is driven by a combination of factors:
1. The impact scales with ambition, problem size, and horizon. A question with a paragraph answer barely stresses the system. You need a reservoir of big, difficult problems that you really care about. This aspect drives the consumer / professional dimension of the funnel.
2. The jaggedness of the system (which I have written about a lot separately). Capability peaks in domains that are digital, verifiable and economically valuable. This is because LLM capability emerges from reinforcement learning on verifiable rewards on a curated environment mixture driven by revenue potential. This aspect primarily drives the area (e.g. math&code) dimension of the funnel.
3. Access. Free-tier, paid-tier, internal.
So this is the weirdness of the moment. The general public has mostly not interacted with these systems. When they have, it looks like a derpy chatbot. The majority of professionals still see only a modest uplift. And a small sliver of professionals are experiencing the vertigo of the curve going vertical. And it is all happening at the same time.
itsbdell 🔎 retweeted
I wrote this thesis for @every almost a year ago; interesting to see how these categories have played out:
- knowledge compounder --> RL / continuous learning / data companies $$
- workflow --> n8n, comfy, eleven labs
- reality gatekeepers --> fintech (stripe, ramp), cybersecurity, compliance (vanta)
- marketplaces --> openrouter, composio, personal agents (soon™)
- vertical transformers --> legora, openEvidence, rogo
the best companies are going to compound multiple of the above advantages all at once, and for these businesses, it gets more moated as intelligence gets cheaper.
itsbdell 🔎 retweeted
Tired: Two-pizza teams
Wired: Two-slice teams
@danshipper on the best team structure in the AI-era
itsbdell 🔎 retweeted
Working With Agents in Slack teapawt.org/i/broadcasts/1qGoNYyeY…
itsbdell 🔎 retweeted
Come work with us and own this product! This idea came out of @danshipper 's brief to me to find a way to quantify our vibe checks of new models from the frontier labs.
I'm working with 15 people so far internally to build their own personal benchmark including @ArielleShipper who runs ops and was our guinea pig for improving onboarding and @kieranklaassen who created compound engineering and spends billions of tokens per day!
You'll work with @nityeshaga who came over to evals with me from the consulting team, where he built workflows and skills for hedge funds, and before that was an early engineer with Kieran on Cora.
You also have support from @JannikJung who helped rescue our vibe coded markdown editor product then automated our internal AI editor KatePass, a clone of Kate Lee our editor in chief.
On design you have @TylerNishida who helped design much of our new @every agent product and had his projects highlighted by the Anthropic team in a new model release.
You'd be working with @bigwilliestyle our head of platform, who just launched the Every agent and previously ran AI operations at Lyft.
I'll be helping you push things forward with the crazy experiments I'm running as head of Evals, but you'll own the product direction as this turns into something real.
Dan is personally very invested in this product and so you get to work a lot with a CEO that actually dogfoods the products that we build (and is spending thousands of dollars in tokens per week himself).
If you have run your own company before (or plan to) this would be a great role for you. Ideally you're a quant that hasn't lost sight of the value of qual (or vice versa). Working at Every is an excuse to get paid to do the fun stuff with AI you would do anyway for free.
NEW JOB ALERT:
@every we've built a personal benchmarking platform, Checks, to help anyone measure how good new models are for their real work
we're looking for someone who understands why this is so important, and wants to help us build it into the future of how great human work gets done with AI:
modern-ton-234.notion.site/d…
itsbdell 🔎 retweeted
I'm super proud of this. Our first conference ever and we've got some amazing sponsors, with a few spots left.
Every’s first conference on the future of work, Thesis: 2027, is coming to @PioneerWorks_ in Brooklyn on November 5. We’re bringing founders, executives, builders, and creatives together to ask what working alongside AI will look like in the next year and beyond.
We couldn’t make this happen without the support of partners who also believe in that future and are actively helping to build it.
Our Founding Partners, @Lovable and @NotionHQ, are two companies that have been a critical part of Every’s journey.
Lovable is the entry point to AI for many of us and is now the software creation platform that gives people the power to act on the problems closest to them.
Notion is the flexible, living context layer behind the people, ideas, and work that have made Thesis: 2027 possible.
Our Presenting Partners are @claudeai, @figma, and @OdeDevs.
Claude, from the team at @AnthropicAI, joins us in bringing together the people exploring and shaping the future of work with AI.
Figma sits at the intersection of design, product, and culture—helping shape not just what we make, but how we work.
Ode with Anthropic helps non-AI companies transform around AI.
Lastly, our Experience Partner, @descript, makes creating videos, podcasts, and clips with AI as easy as editing text.
Thank you to every partner helping us bring this gathering to Brooklyn. We’re almost sold out, but we have a few tickets left. Explore the program and meet the speakers: every.to/thesis-2027?utm_sou…
See you soon.
itsbdell 🔎 retweeted
After automation, you'll give your Legos away to agents
I first learned to give away my Legos, a concept Molly Graham from First Round coined, in 2020. Giving away your Legos is the idea that in order to scale yourself, you have to willingly hand off pieces of your job to new hires and resist the instinct to cling to what you have built.
Back then I was running my first startup and we’d grown to nearly 100 people. I had been used to knowing everything and touching every decision. That stopped working after our headcount passed 50. I had to hand off whole pieces—not just tasks, but decisions. It was uncomfortable, but it was also the only way we could scale.
Six years later, I’m learning to give away my Legos again—except this time, I’m not giving them to a team of people. I’m giving them to a team of agents...
Read the rest here --> every.to/thesis-statements/b…
There are few heuristic, perspective or knowledge breakthroughs more valuable than getting people to go through the obstacle course of doing a new thing
Cc @garychou
Hey friends, I need your wisdom. I’ll be reaching out to some of you individually, but this is an open call:
In about 3 weeks I’ll be presenting in front of ~50 professors of media, marketing, advertising, to give them some guidance / advice / inspo on how they should adjust their curricula and set their students up for success. I’m taking this opportunity super seriously because… AI.
I used to give this talk annually at CES, but this is the first time I’ll be doing it since ~2022, and so much has changed.
My ask: what advice would you give? Not to the students themselves but to those who are teaching and guiding them. Can be anything pithy, but if you’ve also written or tweeted etc about topic, I’d love to know.