Ventry

@elg_oleksandr agreed

@ventry089 pretraining is exhausted. The curve that justified the estimates bends into a plateau.

every "bigger model" launch is a fundraising event dressed as a research one. ilya sutskever - the man who built the age of scaling - just called it. data is finite. pre-training runs out. the curve that justified the valuations is bending flat. but a plateau doesn't close a round. so the pitch gets louder while the architect quietly says the era is over. watch the money, not the leaderboard.

@qyromat0 thanks

profile polymarket.com/profile/0xf5fa…

Sounds insane but one wallet made $1.74M on polymarket - just betting tennis (+7,250x ROI on best positions) 24 bets. all tennis. all wimbledon 2026 here's what i found: palegrit. joined june 2026. zero markets outside tennis. no crypto, no politics, no weather - just balls over a net and he didn't just bet - he sniped: - entered wimbledon markets before lines moved - took underdogs at 3-8 cents - closed when the market caught up 24 predictions. net profit: $1,749,103.90 for context - the average polymarket user makes hundreds of bets and bleeds out. this guy entered 24 times and walked out a millionaire what even is this? insider info? a model? a guy who knows every player personally? his profile is public - you can check every single position. all green the wimbledon sniper. 24 trades. $1.74M. and not a single miss outside tennis - because he doesn't play outside tennis if you don't have an edge - don't enter. and if you do - pick one market and hammer it

@qyromat0 alpha bro thanks

A superagentic AI engineer just showed how to fix the biggest problem with coding agents - in 17 minutes large codebases break every coding agent today. they lose the architecture, drown in tool output, and hallucinate as context grows. recursive language models (RLM) from an MIT paper flip this - the model loads the entire repo into a programmable REPL, writes its own inspection code, and recursively delegates focused sub-questions through llm_query calls. instead of stuffing everything into one context window, it breaks the problem into smaller scoped queries that each see only what they need. live demo runs end to end on both local and cloud models with a fully inspectable trajectory - you can see every step the agent took and why.

everyone's about to sell you a "trading agent." i built one this weekend. the agent is the easy part - anyone can auto-place orders. the alpha is WHO you copy. so i pulled every top polymarket sports trader and ranked them - not by profit (that's size + luck), but by win rate vs entry price. most "$10M" whales just farm 97c favorites. zero edge. the real ones win 60% entering at coinflip prices. that's skill you can actually copy. backtested it: +8% on turnover after slippage. the ones i filtered out lose your money faster - now with an agent. the bot isn't the moat. the filter is. bookmark this before you hand a wallet to a "trading agent."

@Polystreetdotai agreed

@ventry089 Polystreet agents are the best!

$18,644,258 in profit. one wallet. the biggest sports trader on all of polymarket. and he doesn't make it on pre-game predictions. he trades live, mid-match. here's the edge i see every day breaking down wallets like this. a live sports market during the game isn't probability. it's fear. a team concedes - the crowd panics and dumps at any price. momentum shifts - everyone piles into "it's already over". the price flies on emotion, not on real odds. swisstony sits on the other side of that fear. he's not guessing the outcome. he's catching the moment the market overpriced the panic. his biggest open position right now: $144,993 on "france does NOT win on july 14". not a bet on the favorite. a bet AGAINST the crowd that already threw money at the frontrunner. the principle is simple: the louder the emotion in the moment - the fatter the spread between price and reality. and that spread is your edge. $18.6M wasn't built on luck. it was built on the patience to sit against the panic. bookmark this. next time the market is screaming - remember who's on the other side of the trade.

his profile polymarket.com/profile/0x204f…

@vijn_crypto best footballer haha

Polymarket whale just put on BOTH England and Norway to win today He bought around $340K total: > England Yes: 440,991 shares at 50.2¢ - around $221K > Norway Yes: 469,304 shares at 25.5¢ - around $119K So in reality, he paid around 75.7¢ combined for “no draw” If England wins - he can make around $101K profit If Norway wins - he can make around $129K profit Sounds like a risk-free hedge? Not really If the match goes to a draw, both positions can get wiped Haaland can absolutely cause problems for England, but England can also answer back So a draw is very real here I found this trader using @PredictParity and I’m definitely going to keep tracking him Smart structure, but definitely not free money

@xlee_istheking gmgm how are u?

gm gm ct happy sunday to yall

Started at $0. 181 trades. ended at $3,582,415. this is the Unacceptable-Crop-Spotlight wallet. i break down books like this every day, and this one is clean. where he printed - live sports. world cup, tennis, a spread of matches. here's the part most people miss. when a match is live, the polymarket book isn't pricing probability. it's pricing fear. team concedes - the crowd dumps at 0.30 while the real odds are still 0.55. favorite drops a set and the price falls on emotion, not on match math. he bought in exactly those panic seconds. not guessing outcomes - catching the gap between the price and what's actually happening on the pitch. 3.5 million from $0 isn't 181 lucky flips. it's 181 times someone else's emotion handed him a price cheaper than the truth. right now his book is near empty with a micro position on roland garros - halys vs zverev. worked it, cashed out, walked. discipline. principle worth saving: on live markets you're not betting on a team. you're betting against the panic of everyone who already did. bookmark this. next live match, watch the price chart the second a goal lands - that's your entry.

his profile polymarket.com/profile/0x5966…

@slash1sol agreed

SPAIN KEEP WINNING BUT THEY STILL HAVE NOT LOOKED LIKE A TEAM THAT DESERVES TO Spain 2-1 Belgium. Into the semis, but look at how they got there. An 88th-minute winner off a goalkeeper error. Both their goals came from rebounds Belgium spilled. 16 shots, and they still needed a mistake to win. Same script as the Portugal game. Late winner, no clean kill, just barely enough. Next up is France -> that is where this run ends imo. France have been the most dominant side of the tournament. Spain are the luckiest team still standing. That gap gets exposed when the margins go thin. My call: France beat Spain, and the final is France vs Argentina. The money leans the same way. France headlines both top final pairings on @1winToken. The France/Argentina final I am calling sits at 21%, right at the top. Spain's best scenario is stuck at 17%. $1.1M in volume, 108,663 bets, and the market is not buying the Spain hype either. Winning ugly works. Until you meet a team that wins clean. Tuesday, Spain get France. Check & Trade it ↓

@antisadh he's got a good set

$60 ESP32-DIV RUNNING M5SHARK V1.5.3 SCANS DMITRI'S HOME MESH ACROSS 3 ANTENNAS AND FEEDS DIAGNOSTICS INTO A LOCAL QWEN 3.6, KILLED $459 A MONTH IN CLAUDE AND CHATGPT FOR EVERY NETWORK QUERY 00:14 dmitri boots the skull-logo handheld to firmware v1.5.3 and swipes past the menu, "esp32-div by cifertech running m5shark, this replaces my flipper zero and my claude bill in one shot" the rig stacks a $60 ESP32-DIV board by CiferTech with a 240x320 color touchscreen, 3 SMA antennas for 2.4 ghz wifi, 915 mhz lora and gps, m5shark firmware v1.5.3 flashed straight from github, powered by an internal 18650 lipo dmitri runs continuous 2.4 ghz spectrum scans of his own home mesh, logs every beacon, deauth event and rssi drop across a 24 hour window into his obsidian vault before the next boot cycle even hits the vault pipes into a qwen 3.6 27b served off a used rtx 3090 in his apartment, the local llm decodes rf patterns, spots weak encryption on his own gear and drafts hardening plans without a single token touching an api $200 chatgpt pro and $200 claude code max died the week the esp32-div came online because every wifi debugging query, encryption math and mesh optimization now routes through €680 of used silicon at zero recurring cost the window is open, follow and bookmark before it closes