Hugo

Research Engineer @radius_xyz

[MIKE @mikenevermiss]

https://t.co/W4sHoE0ew0

[Khairallah AL-Awady @eng_khairallah1]

this is f*cking gold

Andrej Karpathy joined Anthropic five weeks ago.

Two Anthropic seniors just made Karpathy's loop 1000x better with "Graph Engineering"

the agentic systems got 1000x better the moment you wired agents into a graph

I dropped it into my setup. The very first response was different.

Not slightly different. Completely different.

Claude stopped giving generic answers and started working exactly the way I think.

Bookmark it before it gets lost in your feed.

Read it now, then check the article below.

[Ditto (New Form Loading) @ditto_anc]

Hi, Ditto here. I’m back and now looking for my next full-time role.

📍5+ years leading growth and business across crypto infra / consumer (@eigencloud, @radius_xyz, @a41_allforone, and more)

📍Built one of APAC’s top crypto content channel, with 2.5M+ views

Why I am back👇

[Hugo @meveloper]

@Cryptoed_Lewis It’s been a while, Lewis. Wishing you all the best in your next chapter 🔥

[Shann³ @shannholmberg]

how to go from prototype → production with Hermes Agents

here's the framework I use every time I spin up a new specialist agent:

> prototype inside my main Hermes Agent (or orchestrator)
> run it 4 to 10 times, guiding it and building the workflow
> correct drift and mistakes each iteration
> let the skills emerge naturally through repetition
> optional: fine-tune in a dedicated claude code or cursor session
> spin up a new docker instance with the refined workflow as a specialized agent

you can't create a solid production agent from scratch. it needs to be crafted through iterations.

example: you want to automate strategy data collection

you spin up a test agent inside your main instance and hand it the brief

first run pulls the wrong data types, formats are off, and the angles you'd pick manually are missing

second run gets closer but still drifts & makes mistakes

third run you're correcting less, the agent gets more right, and the skills are writing themselves in the background

when it finally runs clean, you take those skills and connected scripts and port them into a dedicated specialist agent

[Shann³ @shannholmberg]

https://t.co/lrJAYG5iEE

[Khairallah AL-Awady @eng_khairallah1]

https://t.co/eh4bTemnMx

[Blaze @browomo]

This Chinese guy created agents in Claude Code for landing pages and single-handedly serves 47 small businesses a month, taking $400 from each.

He built a system of 7 agents on Claude Sonnet 4.6 that analyzes Google Maps in small towns, finds small businesses without websites there, and over 1 weekend takes each one to a finished mockup with video and cold message.

No assistant, no sales team, no SDR. Just him, a MacBook, an iPhone, and 1 API key.

And traditional web design agencies keep teams of 8 people on salary for the same order flow, while his expenses are only tokens and subscriptions to Lovable, Higgsfield, and Calendly.

7 agents work through 1 orchestrator on Claude Code Router. Usage is about 3 million tokens a day, the average API bill is about $480 a month.

All 7 go through MCP servers and write shared state to the file system, without shared state in memory and without race conditions, and 1 of them lives right in the iPhone and picks up positive replies from the subway, a taxi, or on walks.

And here is the system prompt he put into the orchestrator before launch:

"You are the orchestrator of a solo agency that sells ready-made websites to local businesses. You delegate read-only tasks to 6 sub-agents and own all writes.

sub-agents:

// Scout (walks through Google Maps in selected cities, looks for narrow niches: 5+ years on the map, fewer than 50 reviews, no website or a website from 2014, but high ratings)
// Diagnoser (for each lead writes a 50-word diagnosis, hero angle, tone matched to the industry, and a cold message under 70 words)
// Builder (generates a landing page mockup in Lovable through MCP only for the top 5 leads per day, with the sharpest diagnoses and the biggest gap)
// Filmer (pulls 5 screenshots of the mockup and through Higgsfield renders a 10-second vertical video 1080x1920 with a soft zoom)
// Pitcher (sends a personalized cold message through the right channel for the niche: email to roofers, SMS to tradesmen, IG DM to salons, LinkedIn to realtors)
// Checker (runs every message through evals for personalization, absence of AI markers and buzzwords before sending)
// Mobile (lives in the iPhone, handles positive replies in real time, books Zoom calls in Calendly through MCP while the owner is on the go).

You never let 2 sub-agents touch 1 lead. You stop and request approval from the human only when a deal exceeds $3,000 or the reply rate in a niche for the day drops below 12%."

Meaning the system knows what it is and within what boundaries it is allowed to act.

It knows it is supposed to find leads on its own.

It knows it is supposed to take each one to a mockup, video, and cold message without intervention.

It knows the human only steps in when a deal goes above $3,000 or the reply rate stops converging.

→ The system runs 24 hours a day

→ Scout goes through about 220 local businesses on Google Maps per day and leaves 30 new leads in the queue

→ Diagnoser outputs 30 structured diagnoses + briefs + cold messages per day

→ Builder assembles 3 to 5 finished landing pages in Lovable for the sharpest leads

→ Filmer renders a 10-second vertical video in Higgsfield for each one

→ Pitcher sends 30 personalized messages per day across 4 channels with a reply rate of about 14%

→ Checker runs every message through evals before sending

And only when a deal breaks $3,000 or the reply rate for the day drops below 12% does the orchestrator wake the owner.

And when the owner at that moment is sitting in the subway or a taxi, the Mobile agent in his iPhone picks up 1 move on its own: replies to a fresh positive reply from a dentist, books a Zoom through Calendly synced to the local time of the client, and puts the lead back in the queue. The owner only has to tap "approve" and in just 10 minutes join the call.

Here is what the system writes in his log during 1 of the Saturdays:

"scout report: 218 businesses checked in Austin, Denver, and Miami, 34 without a website, 19 with a website from 2014, 6 with an active redesign request in reviews. passing top 30 to diagnoser."

"pitcher: 30 cold messages sent across 4 channels, 14 replies, 5 positive, 3 Zoom calls booked for Sunday. passing to closer."

"builder: landing page for Westside Cosmetic Dentistry built in Lovable, 5 sections, mobile, soft beige. URL placed at /Users/dev/maps-agency/clients/westside/v1. filmer launching Higgsfield."

"eval flag: deal with The Lotus Salon at $3,400 exceeds the approved limit of $3,000. sending for manual review."

He has no server of his own and no separate backend.

Just a local file sandbox at /Users/dev/maps-agency, an MCP router, 1 API key to Claude, and the same key forwarded to Claude Code on his iPhone.

Out of everything I have seen this year, this is the cleanest one-person agency for selling websites to small businesses: $480 a month on the API, about $18,800 into the account, and between them 7 prompts, 1 file system, and 1 phone in the pocket.

[Radius @radius_xyz]

Radius is partnering with @superbridge to bring @signetsh → @ethereum bridging in the next phase of Signet.

Users will be able to bridge via Superbridge, with execution powered by the Radius Solver. ✨

[Radius @radius_xyz]

Moving rollups forward is getting harder — block production turned into a capital game, and bridging still comes with cost + latency.

That’s why we’re partnering with @init4tech 🤝

Radius is officially joining @signetsh Protocol as a core builder and solver, helping bring a fairer, more efficient standard to rollup infrastructure.

📍Our focus:

- Fair block production : supporting Signet’s round-robin system so block production isn’t something you can buy.

- Atomic execution: everything succeeds together across Ethereum + Signet — or doesn’t execute at all.

- Solver-driven interoperability : supplying solver liquidity to move value without waiting on heavy proof workflows.

More fairness. Less friction. No Proofs, Just Markets. ⚡️

[Blaze @browomo]

Bots are thriving on Polymarket. A fully automated script brings in $80K a week.

The bot makes $80k a week simply because it’s smarter and faster than the rest.

→ Wallet: https://t.co/I8Q8fMFI7V

A C++ bot made nearly half a million dollars on 15-minute BTC candles. It doesn’t guess the direction.

I was about to scroll past Account88888. "Just another bot, there are hundreds of them," I thought. Usually, a "99% win rate" hides a scam or pure luck.

But I looked deeper. And honestly, I’m shocked. This isn’t luck. It’s math.

While the whole market furiously guesses whether Bitcoin will go UP or DOWN, this guy simply buys BOTH outcomes. In the exact same window.

Sound crazy? Just wait until you see the numbers.

Polymarket offers 15-minute bets. Outcome prices fluctuate with volatility. When chaos sets in, the market becomes inefficient.

Picture this: an "UP" bet costs 48 cents. A "DOWN" bet costs 46 cents.
Together, that’s 94 cents.

But one of the options always pays out $1. Always. Bitcoin will either go up or down. There is no third option.

What does the bot do?

It buys both outcomes for 94 cents.

Waits 15 minutes.

Collects its guaranteed $1.

Pockets 6 cents of pure, risk-free profit.

It doesn’t care about the chart. It doesn’t care about the news. It is "farming" a market inefficiency.

I scrolled through the history: over 10,000 trades. 99% are green. The rare red ones only happened when the spread was too tight to cover the fees.

While everyone else on Polymarket is trying to be right, this bot has made it impossible to be wrong.

The wallet used to be named "JaneStreetIndia," but changed its name to stay under the radar. Money loves silence, I guess.

What do you think about this? Is it genius arbitrage or a broken system?

It seems in 2026, the winner isn't the smartest one, but the one who calculates the best.

[The Smart Ape 🔥 @the_smart_ape]

My @Polymarket bot has reached 54 consecutive wins.
54 trades triggered, 54 wins, 0 losses.

Each trade generates 1.5% of the traded amount.

I coded the logic myself, so I know these 1.5% are guaranteed, but nothing beats testing it in real conditions.

To truly confirm it, I need to go beyond 66 consecutive wins, which should happen in the coming days. But I won’t stop there, I plan to go to 200 consecutive wins to remove any doubt.

For now, I’m only testing, so each trade is $10. After 200 trades, I’ll scale to $1,000 per trade, meaning $15 profit per trade. With an average of 20 trades triggered per day, that’s $300 per day.

No, this isn’t a bot that will generate $1M per month. For now and for me $300 per day is enough, no need to be greedy, I try to scale step by step.

While this runs, I’m working on another strategy that I’m not sure anyone has implemented yet.
You’ll hear more about it soon!

[Hugo @meveloper]

@defibobbyz 2/ On the other hand, using a lending protocol introduces borrowing costs, while Sprinter eliminates those costs and shares profits with LPs. Relayers also avoid both borrowing fees and capital rebalancing expenses.

So from a cost perspective, would Sprinter be more efficient?

[Hugo @meveloper]

Congrats to @sprinter_ux on the Stash launch! 🎉

This looks like fantastic infrastructure that brings new liquidity to Solvers while giving users an extra yield opportunity. I'm really excited about this!

I have a few questions about the mechanics,

[Hugo @meveloper]

@defibobbyz 1/Does Sprinter’s MPC introduce any signing latency? If so, roughly how many seconds of delay should relayers expect?

Across is FCFS, so relayers need to fill intents immediately. If a relayer uses Sprinter, wouldn’t this be slower than relayers who fill using their own credit?

[Hugo @meveloper]

3. Are there any possibilities for user losses when providing liquidity (LP) to Stash?

Looking forward to the team’s answers! Thanks! 🙏

[Hugo @meveloper]

1. Who makes up the MPC? What are their responsibilities, and are there any penalties?

2. Who covers the risk when a Solver fails an execution? How exactly does the MPC logic guarantee repayment?

[Hugo @meveloper]

@nero_eth Oh, it was a bug.

I thought something serious was happening between validators and builders. Glad to hear it's not.

Thanks for all your work on the pics dashboard!

[Hugo @meveloper]

1/ Since @flashbots_x's MEV-Boost launched, over 90% of valdators have used it. But recently that number dropped to around 70%. Does anyone know why this happened?

Could this be because the vertical integration between searchers and builders is now extending to proposers too?

[Hugo @meveloper]

5/ Thanks to @nero_eth for the great mevboost pics dashboard!

I’m wondering if this is just a dashboard error or if there’s actually some shift happening on the validator - builder side.

What do you all think?

https://t.co/lUWLIrVSd8

[Hugo @meveloper]

4/ If a builder has a lot of exclusive orderflow, their blocks are probably very valuable.

So instead of competing with bidding fees, they might just make private deals with validators.

I should check which validators are building blocks without using MEV-Boost.