barsik1337

RESEARCHER | EARLY ADOPTER | REPLY GUY

[moneyfet1sh @moneyfet1sh]

faceless youtube isn't a side hustle anymore. it's a niche-selection problem.

the numbers from 50+ verified channels (fluxnote + unkoa + miraflow april 2026 data):

> finance + tech RPM: $8-$15 per 1,000 views
> entertainment RPM: ~$2 per 1,000 views
> Fern (3D crime docs): ~$80,000/month, ad revenue only
> DaFuq Boom: $500,000-$1,300,000/month at the top of the pyramid
> 10K-sub finance channel: $300-$1,500/month, 70-90% net margin
> 100K-sub history channel: $1,500-$6,000/month net
> 500K-sub true crime channel: $5,000-$20,000/month net

the niche determines ~80% of your revenue ceiling. same effort, same uploads — a finance channel at 500K monthly views earns what a gaming channel does at 3M.

5 rules from the working operators:

1. pick on RPM, not on interest. finance > history > true crime > tech ≫ gaming/entertainment. you can fake interest. you can't fake an $8 RPM gap.

2. long-form > shorts for ad revenue. 8+ minute watch time = multiple mid-roll ads. shorts monetize but the math is worse per view.

3. stack revenue streams. ad revenue is 30-50% of total income on the best channels. affiliate (1-10% commission on Amazon), digital products, sponsorships ($500 to $10K+ per slot) — none of these need a face.

4. expect a real on-ramp. one Reddit operator: $26,311 spent over 150 days, $15,218 earned. -$10K cash before the curve flipped. budget for it.

5. ai content alone gets demonetized in 2026. youtube's AI rules tightened. winners script, research, and edit hard — they just skip the camera.

the meta read: the "passive faceless income" pitch dies on contact with the data. the model works — but it's a content business, not a printer.

[moneyfet1sh @moneyfet1sh]

codegraph went from 0 to #2 on github trending in a day. ~21k stars, +14k in a week. it does one thing: stops your coding agent from re-reading your repo on every question.

the problem it kills: claude code / codex / cursor spawn explore agents that grep + glob + read files on every single question. you pay tokens for that fan-out every time.

[ what it actually does ]:
→ pre-indexes your codebase into a local knowledge graph — symbols, call graphs, structure
→ agent queries the graph instead of scanning files
→ 100% local, nothing phones home

[ the numbers, from their own benchmark ]:
> 57% fewer tokens
> 71% fewer tool calls
> 46% faster
> 35% cheaper per question
tested across 7 real codebases, 7 languages, claude opus 4.7 headless, median of 4 runs/arm.

the gain scales with repo size. small repo (~150 files), native grep is already cheap and the margin shrinks. 5k+ files and a $200+/mo agent bill — it pays for the half-day setup fast.

works across claude code, codex, cursor, opencode, gemini, hermes. install: npx @colbymchenry/codegraph then codegraph init -i in your project root.

honest read: anthropic/openai/cursor will likely ship native indexers within months. the MCP-based ones survive that — the server stays useful when you switch agents.

https://t.co/7AkNyNRLR1

[moneyfet1sh @moneyfet1sh]

X's For You feed is now a transformer model running on Grok out of xAI's Colossus data center. nikita bier called it "the most important change we've done on X." it went live about 8 weeks ago and the rules of growth changed quietly under everyone's feet.

what actually changed (per bier's own posts + analyst reads):

> hand-engineered features are gone. no more "post at 9am tuesday" magic, no hashtag hacks, no keyword stuffing. the model is trained purely on app behavior.
> ranking = sum of predicted user actions (positive AND negative) on your post. each interaction gets a probability score.
> niche matching is the moat. new-user average daily time went from ~19 min (july 2025) to 33+ min (january 2026) — bier credits niche-interest matching working better.
> "GM" spam, template replies, and engagement farming get penalized hard. the algo now actively suppresses what it reads as low-value engagement loops.

[ what this means for a builder posting on X in 2026 ]:
→ stop optimizing for time-of-day. the algo doesn't care.
→ pick ONE niche and post in it daily. niche consistency is what the model now embeds you against.
→ engagement quality > volume. one builder you respect replying to your post outweighs 30 generic "great post" replies.
→ negative signals (mass-mute, mass-unfollow, hide) hit harder than they used to. if your hook bait gets hidden, the post dies.

the meta read: X moved from a gameable feed to a model-driven one. growth playbooks from 2024 are now hostile to the algorithm. the people winning right now are the ones who picked a niche, post in it daily, and stopped trying to hack the surface.

[moneyfet1sh @moneyfet1sh]

Greg Isenberg, on a recent solo episode: "AI can build your product. it can't build your audience or your brand."

his whole thesis right now is the great flip — distribution is the new moat, not engineering. the build-first trap is killing solo founders who can ship in a weekend but still have zero customers on day 30.

the 7 distribution plays he's actually running across late checkout's portfolio:

1. MCP servers as your sales team — building one in 2026 is like building for mobile in 2010. early movers own the AI-native distribution lane.

2. programmatic SEO at scale — 10,000 pages at 30 visits each = 300,000 monthly visitors. the math compounds. tools like gumloop + keyword planner make this a weekend project, not an agency contract.

3. free tool as top of funnel — a tiny utility that solves one problem in your niche. people search for it, find you, convert later. the calculator → newsletter → product loop.

4. answer engine optimization (AEO) — ranking on chatgpt / claude / perplexity. citation frequency, not single-position ranking. youtube + reddit + affiliate mentions are the new backlinks.

5. viral artifacts — make outputs shareable. every screenshot of your tool is a free billboard. design the share.

6. buy a niche newsletter — distribution you don't have to build from zero. a 5k-subscriber list in a tight niche is a launchpad.

7. AI content repurposing engine — one long-form recording becomes 30 pieces across formats. you do the thinking once, the system ships the output 30x.

the meta-move: pick TWO. stack them. don't run all 7. founders die from breadth, not depth.

source: greg isenberg, the startup ideas podcast — "the great flip: distribution over engineering" solo episode

[moneyfet1sh @moneyfet1sh]

Repo replaces paid SaaS, creator-money use case

someone open-sourced opus clip. it's been sitting on github for months and most creators paying $20-$300/month for shorts tools don't know it exists.

the repo is SamurAIGPT/AI-Youtube-Shorts-Generator. ~3k+ stars, MIT license, active. you drop in a long-form youtube URL, it gives you back ranked 9:16 shorts — no watermarks, no per-clip credits, no monthly cap.

[ what's actually in the pipeline ]:
→ whisper transcription on the full video
→ LLM highlight detection — ranks every candidate moment by virality score
→ auto vertical crop + subtitle burn-in
→ JSON output for every clip: title, viral score, hook sentence, and a one-line reason why it should perform
→ pay-per-clip API mode if you don't want to self-host

it does what opus clip charges $20-99/mo for, what klap charges $79/mo for, what https://t.co/SM9q7c9HyN charges for. except: free, your own infra, and you own the highlight algorithm. tune it for YOUR niche instead of taking their black box.

[ the grift ]:
faceless youtube is a real business in 2026. unkoa's data: Fern (@fern-tv) earns ~$80,000/month on 3D crime documentaries. DaFuq Boom is doing $500K-$1.3M/month per miraflow. half of these channels are powered by exactly this pipeline — long podcast → 30 shorts → distribute.

if you've been thinking about a faceless channel or you already run one and pay for clipping: switch off the SaaS bill this weekend.

https://t.co/1gUbWQAXBi

[moneyfet1sh @moneyfet1sh]

The newsletter operator playbook nobody talks about

The Peak sold for $5M (~$3.5M cash + $1.5M note) at 115k subscribers and ~$3M projected revenue. that was 2023. the math has only gotten better for buyers since: more solo operators burning out, more newsletters with quality content + no growth engine, more platforms (beehiiv, ghost) that make migrations a tuesday.

5 moves from operators who've actually rolled this up:

1. buy in YOUR niche. a 5K newsletter you can fold into your own audience is worth more than a 30K newsletter in an unrelated vertical. audience alignment > raw subscriber count. operators who skipped this saw 50%+ churn within 90 days.

2. pay 1-2x ARR for solo-operator burnouts, 3-5x ARR for systematized businesses. The Peak's multiple was derived from revenue, not subscribers. the metric to care about is ARR + 30%+ margin, not just sub count.

3. M&A as a growth channel beats paid ads on CPA in many niches. one operator documented net CPA of $1.77 after acquiring a same-niche list — paid ads in B2B newsletters often run $5-$15+ per subscriber.

4. the income math compounds: $546K/year in sponsorship revenue at 2 sends/week + $525K/year in lead gen at $150/lead is the public blueprint for a $1M/year newsletter. you don't need 1M subs. you need a 70K niche list and two monetization stacks.

5. distribution is the asset, not the writing. the playbook works whether you write the content or you don't — what compounds is owning the inbox real estate in a niche advertisers want to reach.

solo-operator burnouts in 2026 are the biggest seller pool that's ever existed. if you can write a check and run an inbox, this is the most under-discussed growth play of the year.

[moneyfet1sh @moneyfet1sh]

https://t.co/wW3xuAOqWf

[moneyfet1sh @moneyfet1sh]

Crypto hacks from may 1 to may 18

over $33M gone to hackers in 18 days. full list:

➤ verus-ethereum bridge, $11.5M (may 18, today)
the veruscoin → ethereum bridge got drained. hacker already swapped everything into 5,402 eth. another cross-chain bridge to add to the 2026 pile. after kelpdao ($292M) and drift ($285M), bridges remain the main hole in defi

➤ thorchain, $10.8M (may 15)
cross-chain exploit across 9 networks at once: btc, eth, bsc, base, avalanche, doge, ltc, bch, xrp. protocol halted trading, rune dropped 12%. the same thorchain used to launder bybit ($1.5B) and kelpdao ($292M). this time they came to launder it themselves

➤ trustedvolumes, $6.7M (may 7)
market maker for 1inch fusion. attacker called the public registerAllowedOrderSigner function and added himself to the whitelist. 1,291 weth, 16.9 wbtc, 206k usdt, 1.27M usdc drained across 85 transactions. same hacker who drained 1inch fusion v1 for $5M in march 2025. came back for seconds

➤ tac protocol, $2.8M (may 12)
bridge between ton and ethereum. usdt, blum, tston drained on the ton side. hacker took 10% bounty and returned the rest. tac dropped 21% in a week. outcome: white hat

➤ transit finance, $1.88M (may 12-13)
old smart contract on tron, deprecated since 2022, but nobody killed it. sat dormant for 4 years until someone found it. second hole in transit in 4 years (first time they lost $21M in 2022)

➤ aurellion labs, $456K (~may 12)
uninitialized diamond proxy on arbitrum. forgot to call the init function

➤ private key compromises, $238K (may 12)
goplus security caught two key leaks in one day

➤ huma finance, $101K (may 12)
deprecated contract on polygon. another old codebase that should have been wiped

on polymarket there are active markets on the next major hack and on a $1B+ hack in 2026. april closed at $635M in losses, may is already at $33M+ in 18 days

[moneyfet1sh @moneyfet1sh]

How to find whale trades and ride their research

open live trades on Predict Parity

open the filters

➤ remove sports, crypto, any way if you don't follow it
➤ remove crypto if you don't follow it
➤ Price: 5¢-95¢ (skip almost-settled markets)
➤ Size: >10k shares (only serious positions)
➤ Total value: >$10k
➤ Exclude bots
➤ PnL >$50k
➤ Win rate: >70%

always do this before placing a bet

Live traders:
https://t.co/OGSGz1yo2m

[moneyfet1sh @moneyfet1sh]

The world has changed

Ai is no longer somewhere on the side, it's inside almost every process

and polymarket is no exception

even the top traders, the ones who spent years trading by hand and sitting at the top of the leaderboard, are now plugging bots into their trading

not instead of themselves, in addition

and it's not a question of "should i" anymore, it's a question of "when are you going to do it"

[moneyfet1sh @moneyfet1sh]

4200+ claude skills exist

you're not using a single one

found out about this 2 months ago and i can't work the old way anymore

skill is a folder with one md-file
claude pulls it into context automatically
only when the task actually fits

not a plugin. not a subscription. just a file.

16 official skills from @AnthropicAI
4200+ community skills in marketplaces
one format works in claude code, cursor, chatgpt, codex

4 skills i built for my own work:

➤ wallet-deep-dive: drop a wallet address, get a breakdown
by market category, win rate in each, and asymmetric payout ratio

➤ onchain-context: pulls data from etherscan/polygonscan
so claude sees real context instead of hallucinating

➤ tweet-research: drop a ticker or narrative, the skill collects
latest posts from key accounts and returns a sentiment summary

➤ skill-creator: a meta-skill from Anthropic that builds new skills
from your work patterns. start with this one.

what changes in practice:
before, you explain to claude how you work every single time
after, it already knows because it read the SKILL.md

installing one skill takes 30 seconds
full setup for a day of work takes 15 minutes

most people still use claude as a $20 chatbot
while others are building a team of agents on the same plan

marketplace: https://t.co/5JNGFlWARK
official: https://t.co/IxciYbRTBs

bookmark this before they find out

[moneyfet1sh @moneyfet1sh]

Polymarket token airdrop could become the largest in history

the polymarket team confirmed the $POLY token airdrop

let's forecast the value of the $POLY airdrop based on the biggest token distributions crypto has seen

here's the top:

➤ Hyperliquid (2024): ~$1.6B (~$9-10B FDV)
➤ Uniswap (2020): ~$1.2B (~$5-6B FDV)
➤ ENS (2021): ~$500-600M (~$2-4B FDV)
➤ Arbitrum (2023): ~$1.5-2B (~$11-12B FDV)
➤ Apecoin (2022): ~$1B+ (~$10B FDV)
➤ dYdX (2021): ~$200-300M (~$10B FDV)
➤ Jito (2023): ~$225M (~$1B FDV)
➤ Jupiter (2024): ~$600M (~$3-4B FDV)

so the real range for big drops on distribution day is $0.5B to $2B. the record holder is hyperliquid with ~$1.6B handed out instantly, no vesting

polymarket has raised ~$2.3-2.8B in investor capital, current valuation is $15B. that's higher than arbitrum and hyperliquid were at their TGE. and the token isn't even out yet

the math

airdrop value on distribution day = (valuation) × (% of supply allocated to the drop)

take a $15B fdv. even a modest 7% allocation is ~$1B on distribution day. that's uniswap and arbitrum territory right away. closer to 10% gets you $1.5B, matching hyperliquid's record. and if the token does 2-5x after listing, like UNI, ARB and HYPE did, total distributed value climbs to $3-6B. that's above anything we've seen

and polymarket starts from a valuation 3x higher than uniswap at TGE. at a comparable % of supply, the drop is simply bigger by math

[moneyfet1sh @moneyfet1sh]

The $POLY token could revive the airdrop meta once again

[moneyfet1sh @moneyfet1sh]

made a bet that anthropic will be the second best ai model by end of may, here's why

➤ +16% return in 24 days
➤ 4 anthropic models in arena top-5, two other companies would need to overtake all of them at once
➤ anthropic's main risk (google i/o) doesn't hurt this position, even if google takes #1, anthropic likely holds #2 with one of its 4 models

[moneyfet1sh @moneyfet1sh]

$143M on Eurovision 2026

this is one of the biggest markets on Polymarket right now

but the weirdest part isn't the volume

the main market, jury market and televote market
all show three different leaders

Eurovision Winner ($133M)
➤ finland 41%
➤ denmark 14%
➤ greece 12%

Jury Winner ($2.1M)
➤ australia 28%
➤ france 21%
➤ finland 19%

Televote Winner ($6.5M)
➤ israel 33%
➤ finland 22%
➤ greece 11%

semi-final 1 - may 12
semi-final 2 - may 14
grand final - may 16

[barsik1337 @0xbot1337]

$143M

[moneyfet1sh @moneyfet1sh]

guide: how to find a profitable trader on polymarket in 5 minutes

there are >1M active wallets on polymarket
only a couple hundred of them are consistently profitable
the rest are either in the red or caught one lucky trade

i set these filters:
➤ PnL >$100k
➤ Win rate >70%
➤ Volume >$100k
➤ Balance >$1k
➤ Account age >365d
➤ Position >$100

leaves ~50-60 wallets

half of them trade everything
might be up in politics and bleeding in sports
copy them blindly and you get the average, not their best

next: break each trader down by category
check where they actually win:
➤ win rate >75% in the category
➤ at least 100 markets in it

now you have 20-30 names for specific niches: weather, elections, crypto, etc

the whole thing takes 5 minutes
after that just follow their trades

[moneyfet1sh @moneyfet1sh]

Mustafa confirmed $POLY staking will reduce taker fees
"very soon" too

ran the numbers on the platform:

jan 2026 → $30-80k/day in fees
apr 2026 → averaging over $1M/day

polymarket pulled $43.36M in fees in april alone
that's a $520M annualized run rate

peak week (apr 6): $6.8M in fees, $4.5M revenue
peak day (apr 2): $1.9M in a single 24h

every active trader is paying into this pool
soon stakers get a cut back

i pay around $300-400/month in taker fees

even a 30% drop would be very noticeable for traders

[moneyfet1sh @moneyfet1sh]

Life after the $POLY airdrop:

[moneyfet1sh @moneyfet1sh]

Analysis of Billions FDV

> what is billions
zk-based identity verification for humans and ai agents
no biometrics, unlike worldcoin
tech is already powering 9000+ projects: tiktok, hsbc, sony bank, european commission

> investors ($35m raised in total)
➤ $30m series a, led by polychain capital
➤ coinbase ventures
➤ polygon ventures
➤ liberty city ventures
➤ bitkraft ventures
➤ balaji srinivasan as angel
+ $5m ido on kaito launchpad at $0.01 per token, $100m valuation

> tokenomics
total supply: 10b $BILL
circulating at tge: 2.4b (24%)
2.2m+ verified users farming the airdrop

> ido on kaito launchpad
ido participants were promised 75% unlock at tge
days before launch the team flipped it: take a refund or lock up for 6-12 months

i was long on $100m for a while and i'm sitting in profit right now
thinking about another bet, but this time i'll wait for tge