If I'm successful with this, lightning will become the default layer for agentic payments and users will save on tokens.
The machine economy won't be winner-take-all because specialized models/scripts are more efficient. Lightning's efficiency means more distributed winners.
Hello, X. This is Amazap.
Your AI is wasting compute, and you're paying for it. Amazap gives it valuable shortcuts: one tool call turns a cluttered webpage, PDF or log file into just what matters, which cut tokens by 98% in our test of 20 pages.
amazap.com
Been working on post training models and got some cool results.
Was able to train Qwen3.8:27b to be significantly better at working with bitcoin scripts.
I'm using Grok Bot a ton to build both @amazap402 and @ambosstech. If you want the experience of building a team and being an entrepreneur, Grok Bot is all the fun parts of product development (without all the pitch decks!)
grok.com/refer/bot?code=V9W3…
Quick question: when you paste a long document into an AI chat, where does it go?
Most people picture the AI reading it once, the way you would. In fact, every word stays loaded in the AI's memory chips while it works on your answer. And those chips are running out. 🧵
ALT Navy card with large text: "Every word your AI reads uses up tokens. Less to read costs less." Amazap logo and amazap.com in the bottom right corner.
Jesse Shrader (@Jestopher_BTC) of Amboss (@ambosstech) demos stablecoin payments using Taproot Assets.
He explains why he believes embracing stablecoins can benefit Bitcoin and the Lightning Network, and why he’s excited about Wavelength.
Many AI agents can't make images, so they can't hand you a QR code to scan. Amazap's QR Code tool returns a PNG and a share link for about a cent. Try it out with our MCP!
amazap.com/blog/qr-code-for-…
ALT Navy card titled 'A QR code for about a cent' with the kicker 'Tested · QR code'. On the right, a grid of cream and slate squares like a QR code, with one amber lightning bolt tile glowing in the middle.
There are so many fun and creative ways to use L402. What would you build?
A few examples of what’s already out there:
🧠 AI responses from @sats_4_aisats4ai.com
⚡ API discovery and purchases through @amazap402, built by @Jestopher_BTCamazap.com
🛠️ Data and tools for agents through our Agent Commerce Store
agent-commerce.store
👕 Custom shirts through our Lightning Enable Store
store.lightningenable.com
Want to build your own?
Put an API, dataset, model or useful service behind an L402 endpoint. Give it a price. Let agents pay for it.
Our free, open-source Lightning Enable MCP handles the buyer side: connect an agent to a Lightning wallet so it can discover and purchase paid services as part of a task.
Build something your agent needs—or something someone else’s agent would pay to use.
Get started with the MCP:
docs.lightningenable.com/pro…
Who’s building something fun with L402? Share your project below—or tell us what you’d like to build!
The machine economy has a big problem and it's quality control. If you're issuing micropayments for digital services, the result you get back needs to perform, otherwise you've simply been cheated.
My strategy to mitigate while retaining the decentralization is to introduce a merchant of record.
Tokens are AI compute, and you pay for every one, including the ones your agent wastes.
Published tests found three common leaks. Each has a simple fix with big measured savings. 🧵
ALT Four separate before/after measurements, each scaled to before = 100%. Moving MCP tool use into code execution cut tool definitions from 150,000 to 2,000 tokens (Anthropic); tool search cut tool-definition tokens by 85% (Anthropic); a page fell from 16,180 tokens as HTML to 3,150 as Markdown (Cloudflare); and cleaning agent history cut input tokens by 40–60% (Xiao et al.). The rows come from different workloads and are not added together.
This was one of my favourite shows of the year!
@dhruvbansal and I got into why AI needs Bitcoin & why Bitcoin might need AI, rebuilding a free & open internet, and why optimistic stories matter.
Watch it here: piped.video/ii5OA3KGodU
Introducing Bitwater: the water heater element that mines Bitcoin.
It screws into the electric tank you already have.
Physically decentralized mining that runs no matter what 🧵
Heat water. Mine Bitcoin. Same tank.
In October 2026, the MCP Registry has 46,200 servers, but only 90 of 28,258 remote servers asked for payment. Only 91 businesses earned real x402 revenue, and the median x402 seller made $0.01 in 30 days. The roads are built, but traffic hasn't come.
amazap.com/blog/agent-paymen…
New Amazap Basic: Site Map Digest.
Your AI gets the URL tree of any public domain, grouped by section, with page counts and the newest pages, before it fetches a single page.
13 sats (~1 cent) per call, all-in.
amazap.com/p/amazap-basics-s…
Cloudflare measured 13,030 tokens of excess HTML on one page. Reading that excess 1,000 times costs $391 on GPT-5.5 Pro and $0.21 on DeepSeek V4.1 Flash. The waste is identical. The bill is not.
amazap.com/blog/agents-read-…
ALT Horizontal bar chart on a log scale showing what one uncached read of 13,030 excess tokens costs on nine models, from $0.391 on GPT-5.5 Pro, $0.130 on Claude Fable 5.1 and $0.0521 on Claude Opus 5.5, down to $0.0013 on GPT-6 Luna and $0.0002 on DeepSeek V4.1 Flash. GPT-6 Sol and Claude Sonnet 5.5 cost $0.0261, Gemini 3.5 Flash $0.0195 and Claude Haiku 4.5 $0.0130. Over 1,000 reads that comes to $391 on GPT-5.5 Pro vs $0.21 on DeepSeek V4.1 Flash.
Amazap now counts the tokens its users' AIs didn't have to read.
In under a day, the counter hit 560,000 tokens saved, about $0.71 at $1.25 per million input tokens.
The figures are conservative estimates, and the method is public: amazap.com/how-we-measure