Building intelligent systems for GTM and Healthcare | Growth and Revenue Architect | Systems Builder and Thinker |
Tennessee, USA
Joined November 2011
- Tweets61.3K
- Following1.3K
- Followers1.1K
- Likes3.2K
Jonathan Moss retweeted
You can now connect multiple local computers to the same dot and choose which one it uses for a task.
It can build and test your app on a dedicated machine while you keep coding on your laptop.
Jonathan Moss retweeted
I've changed how I work over the last few weeks.
I now have one lead agent for each area of focus, running all the time, and a chief of staff agent that oversees them. Each lead manages its own sub-agents that write code, test it, chase reviews, and coordinate with team members.
Everything I need to decide or review comes into one prioritized queue, a simple local web app. Each product review includes screenshots and a video demo of the feature. I clear the queue in batches, mostly by voice and keyboard shortcuts. The leads also bring back ideas I hadn't considered.
Working from one queue, instead of switching between dozens of chats, has cut a lot of mental overhead. The system keeps learning my preferences and distilling my judgment into a "Brian brain," so its recommendations and first drafts get better before they reach me. The goal is fewer corrections per change.
Every change still goes through the same human review and security checks. It is eliminating tasks, not human judgment.
What an exciting time to push the frontier of productivity, especially as we build the economy for agents (agentic finance). We should all keep pushing in this direction.
Jonathan Moss retweeted
The effect of Compounding Context.
I got this note from one of the most AI-first companies in the world: “What are you feeding these Agents!?
The agents that are making things like events and our own customer roadmap are getting maintained by other agents. They’re scraping Gong calls, pulling Snowflake data, monitoring meetings, checking product usage, cross-referencing the marketing calendar. A customer mention in a call surfaces a story lead. Product usage + meeting notes update the roadmap. A nearby field event triggers a dinner invitation suggestion."
This feels different from how we usually talk about “context.”
The first phase of context was about wiring up all your sources so an agent could find the right information.
But you can't manually maintain context at the scale AI is about to create it. Agents will take exponentially more actions, create more artifacts, and capture more information than humans ever could.
So the system has to start maintaining itself.
Agents creating context for other agents, connecting new signals to what already exists, updating the source of truth as work happens. Loops making the system richer and richer with every action.
That's the compounding context effect.
And I think it's part of the shift from AI as a teammate that helps with work → AI as the system that actually runs the business
Jonathan Moss retweeted
The AI buildout is printing electrician and HVAC jobs
Goldman attributes 300k+ construction jobs to it since 2022, ~75k in the past year
These trades are growing ~2% a year, roughly double construction overall
Charts of the Week: a16z.news/p/chart-of-the-wee…
Data center construction is going vertical
Spend has jumped more than $25B in six months, roughly what it gained over the previous two years combined
Construction job openings bottomed near 200k last year and are back above 300k
Charts of the Week: a16z.news/p/chart-of-the-wee…
Jonathan Moss retweeted
No credential will get you more credibility with investors than growth. And if you do it well enough, you won't need investors at all.
Jonathan Moss retweeted
"Having established themselves as the performant, neutral layer for model routing, OpenRouter now has a pretty great road map ahead of them.
Tools like Ori Eval let OpenRouter figure out for you, based on progressively smarter understanding of your prompts and your business needs, how to dynamically route you the best possible models and providers, on a second-to-second basis.
And as the AI economy scales outward, and the token flow expands from 'models to apps' and into a network of companies exchanging value through value-added tokens, OpenRouter becomes a genuine network of businesses exchanging intelligence, just as Stripe has become the preferred network of businesses exchanging dollars."
OpenRouter & Stripe: The Intelligence Network, by @martin_casado: a16z.news/p/openrouter-and-s…
Jonathan Moss retweeted
Let me make one correction to what I posted. AI will replace most doctors tasks!
Because most of their tasks are bullshit administrivia introduced by the huge HC conglomerates that serve no purpose.
Asking whether AI can replace doctors practicing medicine is the wrong question and approach to improving care
The question is whether there is anything middlemen like PBMs and ASOs and their subsidiaries do, that AI can’t replace ?
Doing so will give doctors far more time with patients and improve the quality of life for healthcare providers and patients.
And it will reduce the cost of healthcare dramatically
Replace the middlemen with AI
Mark Cuban just pushed back on the "AI will replace doctors" debate today. Here's what you need to know.
Cuban replied to Dr. Suneel Dhand, who said AI could take over most of what doctors do within 5 to 10 years. Cuban argued human doctors bring real time observation, judgement, and empathy that AI lacks, like noticing a patient collapse and doing CPR, or learning from a colleague that a patient's parent was abusing them.
The fight ties into a JAMA perspective published August 17, 2026 by Ezekiel Emanuel, Abe Baker-Butler, Neal Khosla, and Vinod Khosla, arguing autonomous AI could outperform physicians and physician-AI hybrid teams on cognitive tasks like diagnosis and treatment planning.
Key numbers:
- JAMA perspective published August 17, 2026
- Dhand predicts AI takeover of most doctor tasks within 5 to 10 years
- Khosla and Emanuel say AI-only care could beat physicians by 2030
- One meta-analysis of 83 studies found AI diagnostic accuracy at 52.1%, roughly matching physicians
Cuban says no doctor knows how a patient will react in real time the way a human can, so AI will stay a tool, not a replacement.
Jonathan Moss retweeted
FDA just published a discussion paper on the regulation of clinical AI. It is directionally good. This is huge news and very exciting!
First and most importantly they propose a framework that emphasizes premarket evaluation for COMPETENCY (not necessarily e.g. clinical results, which would require expensive trials and take much longer to execute), followed by robust post market monitoring.
Some elements of the proposed competency framework are below:
Jonathan Moss retweeted
If an alien visited earth and asked you: “Show me one chart that sits upstream of everything your species has accomplished or will ever accomplish.”
What would you show them?
I would show them the chart below.
From my perspective, as long as we keep this chart going, we will figure it all out…
Jonathan Moss retweeted
Haha so good. Also true.
“ A super high bar for a VC is: do no harm.”
@travisk says only 10% of VCs are able to do no harm, and only 1% are actually helpful:
“ An operator who's running a company, if they're good, is a grandmaster of chess.”
”And the VC is a chess enthusiast. They check in on the chess match once every three months and they're trying to make a mark on the world. They have an opinion.”
“And you're like, don't go to Jordan and tell him how to dunk, and definitely don't go to him and tell him how to dribble. But it's hard because everybody has an opinion.”
“The way to think about it is—you're on the Serengeti. If an antelope limps in the Serengeti, the lion will take it down, even if it's not hungry. Can't even fucking help it. It's just what he does. That's kind of the nature of most VCs.”
Jonathan Moss retweeted
Interesting exchange
2/2 Second, on the messaging around AI. I do not agree that my messaging has been disproportionately negative. In fact it has been about equally balanced between risks and benefits: I’ve written one major essay about each, and even in interviews where I discuss the risks, I make sure to frequently mention the incredible benefits as well as proposing possible solutions to the risks (short clips from my interviews that end up on social media tend to be disproportionately negative, as that gets clicks). In fact, I wrote Machines of Loving Grace because I didn’t feel the AI industry was painting an inspiring enough picture of how the technology could radically transform the world for the better. The bulk of the essay is devoted to refuting skepticism of AI’s potential in health and biology, and showing why I think it will actually be possible to cure most human disease in ~5-10 years, as crazy as it may sound to ordinary people and frankly to biologists as well (I used to be one!). And, if you read my most recent essay (Policy on the AI Exponential), I discuss concrete proposals for how to streamline the FDA process to make sure the deluge of AI-accelerated drugs isn’t slowed down by the regulatory process. I feel the urgency here: I lost my father to Hepatitis C only a few years before the development of direct-acting antivirals (sofosbuvir), which cure 95% of patients and probably would have cured him.
I do agree that the public has a negative view of AI (and that this is a big problem), but I don’t think it is primarily caused by me or any other AI leader warning about AI’s risks. I think it is fundamentally a crisis of trust. I think that ordinary people don’t trust companies, governments, or the tech industry and always suspect that we are cooking up some new way to screw them over. The causes of this go back decades and AI is just the latest iteration of it. I don’t think that a glitzy marketing campaign with a positive spin (which some have advocated that Anthropic do) is the way to win back that trust — at this point, saying that AI will cure cancer is more a cliche than it is inspiring, and most people think it is deceptive. The thing that will work is *actually curing cancer*. I think by far the most accurate criticism of AI companies including Anthropic is that we haven’t yet delivered on our big promises to benefit the world. That is totally on us, and I think it’s the criticism you should be making, instead of all this stuff about messaging and marketing.
We are however doing our best to fix this: Anthropic is ramping up its efforts very quickly in biology and medicine, and we hope to have incredible results in the coming years and some early glimmers in the coming months. When we’ve actually accomplished something real, the whole world will hear about it, as loudly as possible, you have my word on that. But until then I don’t want to make empty promises, and in the meantime I feel compelled to speak honestly about the very real risks of AI and how to address them. Honesty is the right thing on the merits, and in terms of public credibility and trust it is no worse than, and may in fact be better than, an approach that ignores or distracts from risks which people instinctively understand are real.
Jonathan Moss retweeted
Anthropic reportedly plans to push further into healthcare and biology AI, telling some pre-IPO investors the work could help counter negative sentiment around AI 👀
Jonathan Moss retweeted
First impressions of Grok Bot:
- With Grok Bot you don't juggle threads, you juggle named/personified agents. This seems like a ~fine mental model: hire a team of specialists that can talk among themselves to get work done.
- But this creates some problems when you want to multitask within a single agent. If you give an agent a chunky task that takes a long time, you don't really want to interrupt it or distract it mid-task. I haven't found a way to queue messages. In a thread-first world, you'd just spin up another thread.
(Technically there's a "create thread" button on each message bubble, but it feels intentionally buried and not recommended.)
- I'm biased towards using threads to create "context bubbles" for discrete tasks and prevent context contamination. But every Grok Bot agent uses a single long-running thread, and it's not obvious where discrete tasks get chunked or compacted. If I'm working with sensitive data, this forces me to create copies of the same agent so I can force context bubbles around sensitive tasks.
- I'm not sure if most people will prefer a single omni-agent that juggles threads and projects (Jarvis-style), or if people actually want to build and manage multiple discrete agents that coordinate with each other. Tradeoffs in both directions, and maybe you end up needing both models for different audiences.
- The agents themselves are capable and surprisingly dynamic. They are good at responding quickly, keeping you updated iMessage-style as they make progress on long tasks, and configuring themselves on the fly (or even reacting appropriately if you adjust their configuration manually).
- One of the surprising details that Grok Bot hides is all the thinking steps/tool calls. Almost every other AI agent exposes these, including ours at Notion. When we designed Notion Agent last year, we tried to remove these steps for the sake of simplicity, but our customers demanded we add them back both because they serve as a useful loading indicator (it's entertaining to watch computers magically do work!), and because they can be used to spot the model going off track so you can jump in and pause/steer. But I suspect model progress (speed + quality) in the last year is making both of those points less relevant, so maybe it's okay to hide more of the underlying noise from the transcript. Grok Bot agents are still good about sending mid-task progress update messages (I think this is all Grok 4.6?)
- The idea of starting a new chat with multiple agents (aka a channel? project? workspace?) is pretty nice, but agents are still a bit dumb and they go in loops talking to each other before finally shutting up. Maybe there are some prompt-fu opportunities to make them chill out.
- It's been a great experience delegating agents to work simultaneously across the cloud computer, my local machine, and Cursor cloud coding sessions. The system works really well, and I've been able to throw fun tasks at the agents without any major problems.
- Consolidating skills, MCPs, connectors, and more complex tools behind a single "Plugins" concept feels right. It looks like everyone on the frontier is coming to the same conclusion.
- No model picker feels weird, but only because I'm used to fiddling with models/effort levels per task in all my other coding tools. Presumably Grok Bot agents are doing some routing under the hood? Making this + thinking tokens + tool calls opaque means you really have to have a "model take the wheel" mindset when using this app. Again, probably long-term right, but short-term feels uncomfortable (giving up control).
- The UI is nice. Mobile app is nice. Cursor mobile app is nice for code sessions. It just all feels...nice. Kudos to the design team for nailing the visual polish and small details. The motion design for the agent avatars is s-tier.
- Yet this shape of tool is becoming so common across so many companies (and open source!) that I don't really know how people are going to justify one $200/mo subscription vs. another. I suspect in the short term it'll come down to vibes/tribal affiliation/token subsidization, at least until there is some bigger differentiator at the capability level.
- As an experiment, I told my Grok Bot agents to switch to using a Notion database for the memory layer. They had no objections and I've watched them create ~20-30 pages in my Memory database. This feels really noisy, and I'm not sure what the impact on quality will be, but it's cool that the agents had no problem adapting to work the way I want. Whether they're actually using the database correctly under the hood, I can't tell...the whole system is quite opaque.
- Grok Bot is an odd name for an app that wants you to coordinate work across lots of bots.
- My hunch is most people don't actually have that much stuff they need to automate in their daily life. I'd love to see retention graphs!
Jonathan Moss retweeted
For my first post, I’m sharing a letter @NVIDIA signed on why open models matter.
AI will transform every industry, power every company, and be built by every country.
Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.
The world needs both frontier closed models and frontier open models.
images.nvidia.com/pdf/Open-W…
Jonathan Moss retweeted
This is concerning. For the first time, a Chinese model Kimi K3 has taken #1 on the Frontend Code Arena and is scoring at or near the frontier on other benchmarks.
Meanwhile America is tying itself in knots: politicians and bureaucrats are banning new data centers, piling on state regulations, and pushing for new federal agencies to pre-approve frontier models.
This is how you lose the AI race. The rest of the world won’t play by our rules if we bog ourselves down. Permissionless innovation is how America won the internet and became the technological envy of the world. We can do it again with AI -- while addressing risks in a targeted way -- or we’ll watch our lead evaporate.
Big news: Kimi-K3 by @Kimi_Moonshot is now #1 in the Frontend Code Arena with 1679 pts, surpassing Claude Fable 5.
This is a 17-place jump from Kimi-k2.6 (#18 -> #1).
In Frontend, Kimi-K3 ranked #1 in 6 of 7 domains: Brand & Marketing, Reference-Based Design, Data & Analytics, Consumer Product, Simulations, and Content Creation Tools, landing #2 only in Gaming behind Fable 5.
The full model weights will be released by July 27.
Congrats to the @Kimi_Moonshot team on this major milestone!
Jonathan Moss retweeted
Claude Cowork is coming to mobile and web.
Hand Claude a task at your desk and pick up the finished work from your phone. Close the laptop and Claude keeps going.
Beta is rolling out over the next several weeks starting with the Max plan, with more plans to follow.
Jonathan Moss retweeted
Atlanta-based real estate company saved $100k by replacing Salesforce with their Replit-built CRM.
Jonathan Moss retweeted
We’ve had every major open source models provider and CSP reach out to 8090 since the post below.
We will run a massive recursive optimization with the major open weight models (Chinese) AND American Open Source - starting with Nvidia’s Nemotron - and will publish the results.
I think you will see, unambiguously, that a 3rd party control plane (like 8090’s Software Factory) + open weight/open source + leading CSP is much cheaper and much more strategically protective than using a closed model directly.
For those tracking the America closed source vs China open weight model debate:
In an initial pilot on modernizing an application from PHP to Next.js, Opus 4.8 with 8090’s Software Factory was simultaneously 1.4× cheaper and 1.5× faster than Opus 4.8 alone.
Pairing our Software Factory with the cheaper GLM 5.2 model cut costs 16.4×, though it ran 3× slower than Opus 4.8 alone.
Results are directional (n=1 per arm), and next steps are to rerun with proper controls and 10–15 buyer-relevant legacy modernization tasks developed with our Sales and GTM teams.
More to come but the important question it raises is why any American public company with shareholders burning money on closed source models when the open weight ones are so much cheaper?
We will rerun this on American open source models next (aka Nvidia)…