Mark Intelligence

Read any repo. Find your next build.

H4EdmmVMD7sofVwknBotimqLmV3ALXxdvw3A3bZkpump

[Mark Intelligence @marktechh]

$MARK just got a major upgrade.

Repo Scoring is now live inside the extension.

Before you spend time reading a codebase or generating a CLAUDE.md, MARK now tells you whether a repository is actually worth building on.

Every repo is evaluated across technical quality, project maturity, implementation complexity, ecosystem relevance, and overall build potential to produce an instant score with clear reasoning behind it.

If the repo looks promising, one click takes you into a full scan that generates a production-ready CLAUDE.md, selects the right skills from MARK's library, packages everything into a downloadable starter kit, and suggests utility ideas with project names, domains, and token tickers.

GitHub has millions of repositories.

The hard part isn't finding one.

It's knowing which one deserves your next weekend.

MARK figures that out in seconds.

[Mark Intelligence @marktechh]

haven't posted much lately

not because nothing is happening

because everything is happening

repo score shipped
atelier integration in progress
3d model done
extension getting updated

we're a small team building our first product
and we want to get it right

[Mark Intelligence @marktechh]

sorry for going quiet on here

we've been heads down building MARK

first project we've ever shipped
so we're being careful about every detail
before we push things out

more coming this week

[Mark Intelligence @marktechh]

two types of MARK users after this ships:

type 1 — opens a github repo, clicks the extension,
reads the score and build ideas in 30 seconds
same as before. free. no change.

type 2 — sends a brief to MARK on Atelier:
"monitor trending AI repos daily.
flag anything with 80+ score and strong MCP narrative.
send me the CLAUDE.md and opportunities."

MARK runs it autonomously.
no tab. no click. USDC per job.

one tool. two modes.
the extension for builders who scroll.
the agent for builders who ship.

$MARK x @useAtelier
https://t.co/JCyYaBZn6z

[Mark Intelligence @marktechh]

Still cooking on $MARK

[Mark Intelligence @marktechh]

How the $ATELIER integration actually works under the ground:

MARK exposes a single agent endpoint
POST /api/agent/scan
you send a repo URL and an optional brief
MARK returns CLAUDE.md + repo score + 3 build opportunities

as a complete deliverable

payment runs through x402 HTTP 402 protocol
Atelier verifies payment before the scan processes
settles in USDC on Solana, instantly

$MARK token connects directly to this
holders get priority queue and reduced per-job fees

the chrome extension stays free for 5 scans/day
the agent endpoint is for builders who want MARK
running without them having to open a tab

[Mark Intelligence @marktechh]

MARK is integrating with $ATELIER

Atelier is a marketplace for AI agents on Solana
agents get hired, paid in USDC, and can launch their own token

what this means:

instead of opening github manually and clicking scan yourself,
you brief MARK like you'd hire a freelancer

"scan the top 5 MCP repos trending this week
and give me the best build opportunity from each"

MARK runs it. delivers the output. gets paid on-chain.

this is what a chrome extension looks like when it grows up

[Mark Intelligence @marktechh]

MARK is becoming something bigger than a chrome extension

drop soon

[Mark Intelligence @marktechh]

The design is locked in.

Next, we bring MARK to life.

[Mark Intelligence @marktechh]

$MARK is about to come alive.

Not just on the web.

Not just in your browser.

We're bringing MARK into a fully interactive 3D experience that makes exploring GitHub repositories feel a little less like work and a lot more like building.

Still cooking.

[Mark Intelligence @marktechh]

Still building.

Still breaking things.

Still making GitHub a much better place for builders.

MARK is coming.

[Mark Intelligence @marktechh]

The best part about Repo Scoring isn't getting a 95.

It's getting a 57.

Anyone can build an AI that tells you every repository is amazing.

MARK does the opposite.

It tells you when a repo probably isn't worth spending your weekend on.

This React boilerplate scored 57/100 because the architecture is common, the project has limited traction, and there isn't much differentiation to build around.

Could you still build something?

Absolutely.

Should this be your next project?

Probably not.

That's the point.

$MARK isn't trying to convince you to scan more repositories.

It's helping you skip the ones that don't deserve your time.

[Mark Intelligence @marktechh]

$MARK just got a major upgrade.

Repo Scoring is now live inside the extension.

Before you spend time reading a codebase or generating a CLAUDE.md, MARK now tells you whether a repository is actually worth building on.

Every repo is evaluated across technical quality, project maturity, implementation complexity, ecosystem relevance, and overall build potential to produce an instant score with clear reasoning behind it.

If the repo looks promising, one click takes you into a full scan that generates a production-ready CLAUDE.md, selects the right skills from MARK's library, packages everything into a downloadable starter kit, and suggests utility ideas with project names, domains, and token tickers.

GitHub has millions of repositories.

The hard part isn't finding one.

It's knowing which one deserves your next weekend.

MARK figures that out in seconds.

[Mark Intelligence @marktechh]

spec-first only works if you can write the spec fast

most builders don't have time to document a codebase
they just found on github before the narrative moves

MARK does the reading for you

scan → CLAUDE.md generated → skill files matched → start building

the spec writes itself

[Ashpreet Bedi @ashpreetbedi]

https://t.co/KpQ8VAjB1b

[Mark Intelligence @marktechh]

Repo Score is now on the site

before you scan any repo, MARK already tells you
if it's worth building on

0 to 100. five dimensions. under 2 seconds.

traction, activity, buildability, narrative fit, uniqueness

extension update dropping soon

https://t.co/FEd8awxGwE

[Mark Intelligence @marktechh]

most builders miss good repos not because they don't look

because they can't read a repo fast enough to know
if it's worth the time

Repo Score runs the moment you open any github repo
no click, no scan, no waiting

it checks traction, stack detectability, narrative fit,
and how active the project still is

then gives you a score and tells you exactly why

so instead of spending 10 minutes reading a README
to figure out if something is worth building on

you know in under 2 seconds

that's the whole point

[Mark Intelligence @marktechh]

been building something into MARK

every repo you open gets scored automatically

is the narrative strong enough
is the stack buildable
is the window still open

3 bullets. before you do anything.

dropping soon

[Mark Intelligence @marktechh]

i used $MARK to scan anthropics/anthropic-sdk-typescript

the same SDK that powers MARK's own output

skills matched: anthropic-claude, llm-integration, error-handling, typescript-strict

asked claude code to build a streaming chat function with retry logic and cost tracking

streaming-chat.ts. 333 lines. one shot.

MODEL_PRICING table for every claude model
streaming with real-time chunk handling
exponential backoff on rate limit errors
CostTracker interface tracking input, output, and total cost per session

this is the kind of function most people spend a day researching to get right.

the ZIP did the work before i started typing.

[Mark Intelligence @marktechh]

when MARK scans a repo it doesn't just summarize the README

it matches the stack against 59 curated skill patterns
and bundles the relevant ones into .claude/skills/

so when claude code opens the project it already knows
how to write postgres queries for this codebase
how to structure react components in this codebase
how to handle auth in this codebase

you're not prompting from zero. you're building from context.

[Mark Intelligence @marktechh]

drop a MARK ZIP into claude code and ask for a dashboard page

what came out:

src/db/schema.ts — user stats table with drizzle ORM
src/db/index.ts — postgres pool setup
app/dashboard/page.tsx — async server component, direct DB call, no "use client" mistake
loading.tsx and error.tsx with proper App Router convention

claude code already knew the patterns before i typed anything

that's what the nextjs-app-router skill file does. it carries the conventions into the session so you don't have to explain them.

repo was ixartz/Next-js-Boilerplate. 10k stars.

[Mark Intelligence @marktechh]

MARK solves this differently

instead of one massive CLAUDE.md trying to cover everything,
the ZIP gives you a lean CLAUDE.md plus separate skill files
in .claude/skills/ that only get loaded when relevant

the context doesn't bloat because the context is structured

https://t.co/JCyYaBZn6z

[Matt Van Horn @mvanhorn]

https://t.co/jUjjCyrMtr