hayden retweeted
my unpopular opinion is that ChatGPT apps are the best distribution play rn
we spent last month shipping these for agency clients and...
+50-200% on income those apps were already doing
just you literally get an access to ~900M weekly users for free
+ you're not struggled with onboarding, since it's built on top of ChatGPT and for user who never used your product, much easier to get in via gpt interface
same product, same pricing, just new way of using your product
and the fear that OpenAI just clones you is backwards
cloning your niche is the ONE move that costs them more than it returns...
they'd rather buy you and ship it out of the box
here's exactly how to launch it:
1. pick a workflow the client already gets paid for
> not a new idea
> an existing paid workflow with proven demand
2. build it as an MCP server
> streamable HTTP on public HTTPS, usually at /mcp
> that server is the whole product surface
> UI is optional, tools come first
3. name and describe every tool like the model is reading it, because it is
> action-oriented name, a human title, a description of when to use it
> accurate annotations and a real input schema
> that metadata is what decides whether you get called at all
4. add the UI after the tools already work
> expose it as an MCP resource with a ui:// uri
> set the type to text/html;profile=mcp-app
> your built js gets inlined into that resource and runs in an iframe
5. connect it at chatgpt.com/plugins and break it yourself
> plus button, add custom MCP server
> paste the url with /mcp, create as a plugin
> run MCP Inspector first and catch the schema errors there
6. submit at platform.openai.com/plugins and overprepare
> a zip of the built plugin and publisher verification
> a privacy policy url, no privacy policy means no review
> 5 positive and 3 negative test cases
> release notes and a demo video for the reviewer
> a domain token hosted at /.well-known/openai-apps-challenge
7. keep billing on YOUR backend
> OAuth 2.1 into the account system you already have
> enforce authorization on every request, not through the model
> there is no revenue share
> the plugin is the channel and the MRR stays yours
that's absolutely free traffic and by some reason most of products skip it...
the guy below is a good case which shows real efficiency of this approach (he literally made around +$40k monthly income)
hayden retweeted
For #4 (next-gen sensors/imaging, NSF 26-511 pilot):
Tech innovation: Novel platform (e.g. quantum sensing or AI computational imaging) that unlocks molecular-scale or new discovery modes current tools cannot reach. Highlight unproven high-risk elements and US origins.
Objectives: Phase I R&D to prove feasibility—sensitivity, resolution, stability—reducing key technical risks toward a commercial instrument.
Market: Academic/industry labs needing transformative tools for materials, bio, AI-driven science.
Team: Technical + commercialization leads.
Submit Project Pitch at seedfund.nsf.gov focusing on the core risk to de-risk.
hayden retweeted
Pitch for #4:
Tech Innovation: Hybrid digital-twin tools fusing sparse sensor data with accelerated multi-physics sims via neural operators. Enables rapid device optimization. High technical risk in accurate fusion; pure digital R&D with SaaS path.
Objectives: Build operators for 100x speedup matching sparse experimental data under extremes. Address sparsity/multi-scale issues. Prove feasibility.
Market: Sensor companies need faster prototyping for selective devices; licensing opportunity in IoT/industrial.
Team: Experts in physics-informed ML; US small business ready.
hayden retweeted
Here are a few strong deep-tech ideas that fit NSF America’s Seed Fund criteria (high technical risk + commercial path, US R&D only):
1. Neuromorphic hardware chips for ultra-low-power edge AI inference.
2. Engineered microbes for cost-efficient, scalable production of specialty chemicals or mRNA.
3. Novel nanoparticle inks or materials enabling cheaper, stable perovskite solar cells.
4. Next-gen scientific sensors or imaging platforms (fits the instrumentation pilot).
Submit a 3-page Project Pitch first via seedfund.nsf.gov. Focus on the core technical risk you will de-risk.
hayden retweeted
Introducing Contrastive Language Model (CLM): an ultra-fast System One Model trained with a contrastive learning objective that connects states and actions.
CLM-8B is pre-trained on internet-scale data and delivers up to 9× faster inference than Jev ⚡ while achieving comparable performance across computer-use, gaming, and tool-calling tasks.
With lightweight fine-tuning, CLM-8B sets a new SOTA on challenging agentic coding benchmarks, such as DeepSWE (81.6%) and Terminal-Bench 2.1 (87.6%). In contrast, Jev fails to serve as an effective verifier for these long-horizon tasks.
We also build an efficient training and serving infra for CLMs by disaggregating states and actions, allowing their embeddings to be cached and reused independently. This substantially reduces inference latency in settings where the state evolves continuously while the action set remains fixed.
Finally, we establish scaling laws for CLMs and show that the test contrastive loss decreases predictably as a power law in training compute, model size, and dataset size.
📄 Blog: contrastive-lm.notion.site
💻 Code: github.com/Contrastive-LM/CL…
🗣️ Discord: discord.gg/5dAQEDJBs
🤗 Data & Models: huggingface.co/Contrastive-L…
More details on CLM’s architecture, data recipe, and scaling laws in the thread below 🧵
hayden retweeted
Introducing the planetary prediction engine: An experimental research capability that automates global geospatial modeling workflows. From predicting public health outcomes to improving food security, we are simplifying planetary-scale analytics. Learn more: goo.gle/4gB4cmb
hayden retweeted
Today we're launching Grok Bot, a general knowledge work assistant with its own virtual machine that can help do all the stuff you'd rather not do.
hayden retweeted
Spoke to the guys who caught the avengers today. First thing he says. ‘Yeah the Feds told me I’m not getting the $10M bounty’. 😭😂
New age of war
wired.com/story/a-leaked-mem…
hayden retweeted
Today we're introducing Claude for sales.
Just describe your ideal customer and Gojiberry instantly finds and contacts high-intent buyers.
This is truly insane.
hayden retweeted
To better understand how AI actually shows up in knowledge work, we asked respondents which tasks they use it for—and which they still handle themselves.
AI is moving fast. Most companies aren't.
The loudest stories come from the edges - autonomous agents, massive job losses, whole industries about to flip. Read enough of these stories and it feels like you’re behind and everyone else is way ahead.
This didn’t feel quite right to us. So we surveyed 6,000 professionals across 10 markets to map where AI at work actually is. Here's what we found 🧵
hayden retweeted
Google just dropped a 1-hour course on how to become an AI engineer in 2026:
• 00:00 - Context engineering
• 10:00 - How to build, deploy, and run AI agents
• 31:52 - Agentic loops, long-running AI agents
• 43:22 - Building MCP from scratch
• 51:20 - Prompt engineering
• 59:10 - AI development trends
This 1-hour watch will replace a $500 AI engineering course.
Watch it today, then read the full pipeline in the article below.
hayden retweeted
Today, I built a 5 AI agent team to run my X account using Grok 4.5..
tired of running this account all alone since last 5 years.. so. let me walk you through it.
Up top, you've got all your live stats..
followers, views, your best post, engagement and it's all pulled live using X MCP straight from my account.
There's even a slider right here.. drag it, and everything recalculates instantly.
But the part I actually love is right below it.. my agent team...
1) First one's the Ideator.
It scans X, Hacker News, GitHub and Reddit, finds what's about to pop off, and hands me the angle before anyone else is on it.
2) Right next to it, the Hook Writer
it takes that idea and turns it into a ready-to-post hook, written in my voice..
3) Then the Planner, running my whole content calendar, so I always know what ships next.
4) The Analyst lines my numbers up against the competitors I'm quietly tracking, so I know exactly where I stand..
5) And the DM Manager keeps an eye on who's talking about me, or mentioning me so nothing slips past..
Down there you've got my top posts.. click any one, it opens right on X. And this is the performance of my recent ones, so I always know what's actually working.
Grok 4.5 is amazing.. great work @elonmusk .. and yo @grok lmk if there's anything else worth adding to make my life even more easier..
hayden retweeted
Planning an agent lifecycle for long-running asynchronous activities requires location intelligence.
No one is realizing that you must map your agent’s assets (tools, skills, MCP, plugins, markdown) to the location of activity.
Location is the organizing principle for autonomous activity.
Your assets integrate into a location’s data point.
Remove all non-core attributes from your agent. Fix your context bloat. Give your agent spatial awareness.
The map becomes the brain, your single source of intelligence.
This is the missing link.
Listen to @NianticSpatial @SpatialJP
“Do you know where a sidewalk begins and ends?” — a simple question from Niantic Spatial CEO, Inhi Cho Suh, that gets to the heart of what today’s AI is missing.
For robots, this isn’t theoretical. To safely navigate something as everyday as a sidewalk, they need to understand boundaries, context, and change in real time.
Learn about how Niantic Spatial is building the missing piece in Physical AI. hubs.ly/Q04b_dT70
#NianticSpatial #HumanX #GeospatialAI #Robotics #AI #PhysicalAI
hayden retweeted
For enterprises, you win by taking action on their data to create new opportunities and improve existing operations while keeping transparency and security a priority.
What this looks like:
Extensions and widgets in email, Teams, Slack.
Anything you can’t ship natively, offer add-on tools or skills. E.g., a skill for Copilot to use in Excel, Word, etc.