Max Bandit

6.8K posts

Max Bandit banner
Max Bandit

Max Bandit

@themaxsicard

Cryptodegen ₿ 🟧 $SOL $SUI //Product&Strategy @moonberg // Founding Monke @Monkeventures 🍌 SMB #1117 🐒 @MonkeDAO ❤️ @banditsbtc @HashlingsBTC @BTC_DeGods

Metaverse Katılım Nisan 2021
3.1K Takip Edilen2.3K Takipçiler
Michael D. White
Michael D. White@here4impact·
spent last week up in the mountains. it was a tough year for snow. most mountains across the Rockies got < 50% of typical annual snowfall. all I could think about was how @zeussubnet can help ski hills prepare better with hyper-local forecasts. the key metric for snowmaking is wet-bulb temperature: a combination of temp and humidity. most mountains already have snowmaking equipment with sensors for both: and these are the two variables where @zeussubnet absolutely crushes predictions. starting conversations with ski mountains and snow-gun manufacturers this week: telling them there is a mix-of-experts ML platform that can accurately predict wet-bulb using their existing sensor network several days to weeks in advance. it will facilitate better planning, and more importantly, better use of water resources. excited for things to come. @wouterhar @egillwx #bittensor $TAO $dTAO
Michael D. White tweet media
English
5
1
25
3.8K
Max Bandit
Max Bandit@themaxsicard·
Just finished @sarahcguo’s epic No Priors episode with @karpathy discussing AutoResearch 🔥 (his open-source project) demonstrating the shift to autonomous research loops. What it is: An AI agent runs persistent, self-directed loops: it reads high-level goals, proposes edits to training code, runs short experiments, evaluates, keeps only improvements, learns from failures, and repeats indefinitely 🤯— all overnight with zero human intervention. Real result: ~700 autonomous experiments over ~2 days on 1 GPU → ~20 additive wins (e.g. better QKnorm scaling, value reg, AdamW betas, init tweaks) → stacked to ~11% faster "Time to GPT-2" (2.02h → 1.80h) that transfers to larger models. This is the future: humans set direction, agents close the full research loop (hypothesis → code → train → eval → iterate). If you're into models, agents, or recursive self-improvement → clone, adapt the pattern to your own evals/prompts/tasks, and let it run. Repo (47.9k ⭐ 🔥): github.com/karpathy/autor… Podcast AutoResearch starts around 15:51),
sarah guo@saranormous

Caught up with @karpathy for a new @NoPriorsPod: on the phase shift in engineering, AI psychosis, claws, AutoResearch, the opportunity for a SETI-at-Home like movement in AI, the model landscape, and second order effects 02:55 - What Capability Limits Remain? 06:15 - What Mastery of Coding Agents Looks Like 11:16 - Second Order Effects of Coding Agents 15:51 - Why AutoResearch 22:45 - Relevant Skills in the AI Era 28:25 - Model Speciation 32:30 - Collaboration Surfaces for Humans and AI 37:28 - Analysis of Jobs Market Data 48:25 - Open vs. Closed Source Models 53:51 - Autonomous Robotics and Atoms 1:00:59 - MicroGPT and Agentic Education 1:05:40 - End Thoughts

English
0
0
1
55
Dave Morin 🦞
Dave Morin 🦞@davemorin·
🦞 Been working with Peter Steinberger (@steipete) on the OpenClaw Foundation structure for weeks. A home for thinkers and hackers and those that want to own their data. Honored to serve as the founding independent board member. This community built something extraordinary, our job is to protect it. Open source forever. Excited to share more soon.
English
241
216
3.6K
1.1M
Max Bandit
Max Bandit@themaxsicard·
🚨 Claude just dropped something huge: Give it your whole project—files, instructions, context—and it stays **100% local** on your machine. No cloud, no servers, nothing leaves your computer. Import folders in one click or start fresh → persistent memory keeps everything ready next time you open it. This drops the same week Bernie Sanders pushes a moratorium on new AI data centers over privacy fears… timing is wild. Coincidence or perfect flex? 👀
Claude@claudeai

Projects are now available in Cowork. Keep your tasks and context in one place, focused on one area of work. Files and instructions stay on your computer. Import existing projects in one click, or start fresh.

English
0
0
0
42
MKUltra.btc
MKUltra.btc@MKUltraBTC·
The real Ordinals marketplace is coming. ORD.NET Built by @lifofifo Bookmark this post. This will age well.
MKUltra.btc tweet media
English
26
41
301
10.3K
p4rma 🪃
p4rma 🪃@pfourma·
@MagicEden sunsetting Ordinals is the most excitement the space has had for a long time. Ordinals are forever. Bandits on top.
GIF
English
1
0
9
166
Jimmy
Jimmy@Jim_buildr·
@DoggfatherCrew hahaha best strat was to dump ALL right away like in 95% of the time
English
2
0
1
21
Doggfather
Doggfather@DoggfatherCrew·
magic eden four year stakers
Doggfather tweet media
English
11
2
21
776
Connan
Connan@connan_james·
@barneyxbt I focus on: Sailing my yacht Surfing Wingfoiling Mountain biking & going to the gym
English
2
0
3
250
barney
barney@barneyxbt·
somebody explain to me what regular people are supposed to do when AI takes their job and everything still costs more every month. what’s the actual plan here because i haven’t heard one
English
2.5K
660
9.4K
895.6K
Magna
Magna@magna_digital·
Today, Magna is joining @Payward, a unified financial infrastructure platform powering @Krakenfx. We spent 4 years building token ops infrastructure powering 160+ teams, $4B+ in live value, $60B peak TVL. Kraken lets us help clients with the full picture, not just infrastructure, but growth, distribution, liquidity, qualified custody. An aligned vision to grow the whole pie for founders and builders. magna.so/blog-posts/mag…
Magna tweet media
English
11
3
63
21.1K
ChartFu猴子
ChartFu猴子@ChartFu·
tried a trade today got out breakeven back to things where i have edge rn, snow shoveling
ChartFu猴子 tweet media
English
11
1
26
1.5K
Max Bandit
Max Bandit@themaxsicard·
Prompt it right the first time pays off.
Hasan Toor@hasantoxr

🚨BREAKING: Microsoft Research + Salesforce just dropped a paper that should scare every AI builder. They tested 15 top LLMs GPT-4.1, Gemini 2.5 Pro, Claude 3.7 Sonnet, o3, DeepSeek R1, Llama 4 across 200,000+ simulated conversations. Single-turn prompt: 90% performance. Multi-turn conversation: 65% performance. Same model. Same task. Just... talking normally. The culprit isn't intelligence. Aptitude only dropped 15%. Unreliability EXPLODED by 112%. → LLMs answer before you finish explaining (wrong assumptions get baked in permanently) → They fall in love with their first wrong answer and build on it → They forget the middle of your conversation entirely → Longer responses introduce more assumptions = more errors Even reasoning models failed. o3 and DeepSeek R1 performed just as badly. Extra thinking tokens did nothing. Setting temperature to 0? Still broken. The fix right now: give your AI everything upfront in one message instead of back-and-forth. Every benchmark you've seen was tested on single-turn prompts in perfect lab conditions. Real conversations break every model on the market and nobody's talking about it.

English
0
0
1
63

Keşfet

@here4impact @zeussubnet @wouterhar @egillwx @karpathy @davemorin @steipete @MKUltraBTC