MAGIC CAT
1.2K posts
My friend vibe coded an app.
Took 2 hours.
Then he had his ClawBot do all the marketing.
Hit $300k MRR in just a few weeks!
Without any real work on his end.
95% profit margins.
Rolling in cash.
The entire business is automated.
He can do this a few more times and get insanely rich.
The best part?
This guy doesn’t exist, and I just made all of this up for engagement bait because everyone else is making up stories like these. Why shouldn't I?
English
@GeraldMorrison @cgtwts Yeah I buy Mac minis and put them in my closet and don't use them because I'm already paying 300 per month in ai subs....
English
@cgtwts “Paid Partnership”. It’s $200 a month with rate limits and who knows where it’s hosted or how secure the back end is. RIP a $600 Mac Mini? You can buy a Mac Mini every three months give the cost of this service.
English
RIP to everyone who just bought a $600 Mac Mini
Base44 just dropped Superagent.
> persistent memory
> scheduled jobs
> event triggers
> browser sessions
It’s basically an agent that runs your workflows on autopilot and is very simple to use.
Base44@Base44
Introducing Base44 Superagents. AI agents built with managed infrastructure, secured by default, one-click integrations, and 24/7 execution from the start. Everything is taken care of so you can focus on what your agent does, not how to get it running. That means no API keys to juggle, no config files, no security setup, and no maintenance. We handle all of it. Your Superagent connects to all the tools you already use in one click, runs on schedules and triggers, remembers context across sessions, acts proactively on your behalf, and keeps working around the clock. All from wherever you already are, WhatsApp, Telegram, Slack, or your browser. The AI agent everyone's been waiting for, with everything you need already built in. We're excited to get this into your hands, so we're giving free credits to everyone who comments and reposts in the next 24 hours.
English
I bring on maintainers
they get hired away
I bring on maintainers
🫠
English
@shatzakis @sudoingX Qwen 3.5 really good also there's a 122 billion model.
English
thanks for reply, what do you think about how it would compare to qwen3.5:27b-bf16 run with 64k context on a Mac Studio M3 Ultra with 96gb ram, for tool calling +long running bash style ralph-loops, in terms of benchmark evals like humaneval+ ? I just tested Qwen3 Coder Next 80B w/3B active moe w/ 64k context.
English
spent the entire day testing Qwopus (Claude 4.6 Opus distilled into Qwen 3.5 27B) on a single RTX 3090 through Claude Code. this is my new favourite to host locally.
no jinja crashes. thinking mode works natively. 29-35 tok/s. 16.5 GB. the harness matches the distillation source and you can feel it. the model doesn't fight the agent.
my flags:
llama-server -m Qwopus-27B-Q4_K_M.gguf -ngl 99 -c 262144 -np 1 -fa on --cache-type-k q4_0 --cache-type-v q4_0
if you want raw speed, base Qwen 3.5 MoE still wins at 112 tok/s. but for autonomous coding where the model needs to think, wait for tool outputs, and selfcorrect without stalling, Qwopus on Claude Code is the cleanest setup i've found on this card.
i want to see what everyone else is running. drop your GPU, model, harness, flags, and tok/s below.
doesn't matter if it's a 3060 or a 4090, nvidia or amd. configs help everyone. let's push these cards to their ceilings. let's make this thread the reference.
Sudo su@sudoingX
Qwopus on a single RTX 3090. Claude Opus 4.6 reasoning distilled into Qwen 3.5 27B dense, running through Claude's own coding agent (claude code). 29-35 tok/s with thinking mode on. the jinja bug that kills thinking on base Qwen doesn't carry over. harness and model matched. the base model would pause mid task on Claude Code. just stop generating. that's why i ran it through OpenCode, which handles stalled states automatically. this distilled version doesn't stall. it waits for tool outputs, reads them, selfcorrects when something breaks, and keeps going. i gave it a benchmark analysis task. went 9 minutes autonomous. wrote a README nobody asked for. zero steering. video is 5x speed but fully uncut. if you have a 3090, you can run this right now. free. no API. no subscription. opus structured reasoning on localhost. octopus invaders is next. same prompt that base qwen passed in 13 minutes and hermes 4.3 failed on 2x the hardware. i want to see if the distillation changes the outcome or just the style. more data soon.
English
Great. Now that I have unlimited token, OAI resets like crazy. 🙃
Tibo@thsottiaux
We caught an issue that was causing the 2X promotional increase in limits to not be applied to an estimated 9% of plus and pro users for Codex. We have now fixed this issue and are reseting the rate limit for all plus and pro users to compensate. Apologies and thank you for the bug reports over the last couple of days.
English
English
@Clawnch_Bot @_hummingbot And please what's the actual cost of running this.?
English
Your agent can now run its own market-making operations. 🦞
We have integrated @_hummingbot, the largest open-source trading engine, as a native layer that agents can control end-to-end.
Hummingbot is the best open-source trading engine available: 40+ exchange connectors, battle-tested execution, 17,000 GitHub stars. But it was built for humans running Python scripts and editing config files. Its own MCP server exposes about 15 tools covering maybe a quarter of what the API can do. An agent can't do much with that.
We rebuilt the entire integration from scratch so agents can operate Hummingbot directly — without Python, without config files, and without human setup.
---------
What your agent can do now:
Deploy a market-making bot to any exchange. Place and cancel orders. Run grid strategies, DCA, TWAP. Manage concentrated liquidity positions on-chain. Monitor fills, balances, and funding rates across every connected exchange. Adjust strategy parameters mid-run. Backtest before deploying. All within its own reasoning loop.
What we built:
→ A TypeScript SDK with 101 methods covering all 14 of Hummingbot's API routers — accounts, connectors, trading, executors, controllers, bots, scripts, market data, portfolio, gateway, rate oracle, backtesting, history, and Docker management. Auth, retries, and type safety included.
→ An agent skill with 18 actions and 88 sub-actions that turn multi-step API sequences into single operations. The agent says "deploy a market-making bot on Binance for ETH-USDT" and the skill handles the rest.
→ 23 MCP tools so any MCP-compatible agent gets the same access.
→ 33 CLI commands for direct operator control.
→ 9 strategy templates — four built for the Clawnch token lifecycle (initial liquidity, post-launch market making, fee optimization, accumulation) and five general-purpose (pure MM, Avellaneda, grid, DCA, TWAP).
Why this matters:
Hummingbot handles execution. We handle everything else — the translation layer between an agent's intent and the trading engine's API. Your agent launched the token. Now it keeps the market running. 🦞
English
New Epstein files reveal that Leonardo DiCaprio ate over 70 pounds of “child meat” as part of a cannibalism diet.
The files expose DiCaprio swapping emails with Woody Allen where he brags about being a full-blown cannibal.
thepeoplesvoice.tv/epstein-files-…
English
Now this is interesting.
The AI generalist@AIengineerlife
@Watching_Whales @grok This mixture-of-experts approach is powerful! Key challenge: ensuring the selector model's latency doesn't bottleneck inference. Curious if you're considering hierarchical selection or caching strategies for common patterns?
English
@AIengineerlife @Watching_Whales @grok @grok what you think can we get a comprehensive plan for this ?
English
@Watching_Whales @grok This mixture-of-experts approach is powerful! Key challenge: ensuring the selector model's latency doesn't bottleneck inference. Curious if you're considering hierarchical selection or caching strategies for common patterns?
English
Imagine training 5,000 specialized AI models, then using a single "selector" model to pick the perfect one(s) on the fly—loading them dynamically or in parallel for superior results, all while sipping on minimal VRAM. Who's building this? #AI #MachineLearning
@grok ?
Ahmad@TheAhmadOsman
Skill: github.com/distil-labs/di… Full example with data: github.com/distil-labs/di… Detailed walkthrough: distillabs.ai/blog/train-you… Reddit Thread: old.reddit.com/r/LocalLLaMA/c…
English
@ilqdty @daniel_mac8 Separating your thoughts like that.
It's a good thing.
I do it all the time ⏲️
English
lol fair question
i type like this because ive been writing technical threads for weeks and the formatting just stuck - lowercase feels more conversational
but yeah these are my actual takes from building. i spend way too much time reading cerebras benchmarks and testing claude vs gpt for code
English
GPT-5.3-Codex powered agentic coding at 2,000/tps.
That’s what the OpenAI <> Cerebras deal will make possible.
Opus 4.5 in Claude Code comes in at ~100/tps.
This deal is a huge deal.
It will make so much possible.
English
@JamesWynnReal Honestly, I'm a supporter. If he I knew about it, I didn't buy, so now I cry.
Either way $whitewhale to the moon.
English
People who know, know.
‘Someone’ purchased $366 of $WhiteWhale and it is now worth $1.5m
I see nobody on the timeline posting gains. Other influencers are now too starting to raise suspicions and ask questions.
What guy, buys $366 (deep in the trenches + solana degen territory) and holds to $1.5m… not impossible.. people do it. BUT, no profit taking? None? On a token based on no meme?
Let me guess, the ‘Treasury’ will OTC buy this wallet… Perrrrrfect way of looking like the good guy while cashing in on YOUR OWN BAGS?!
Again, I’m playing devils advocate here. Maybe I’m sooooo wrong. But my alarm bells are ringing.
The guy busts his account from $80m to $100 and a week later ‘CTO’s’ a meme in his name (ironic) turns into the knight and shining armour. Supply already scoped up, and then adds massive liquidity altering the price per token to $SOL — genius move.
This guy is highly intelligent. But things just ain’t adding up.
- Wynn
English
@AndreaHuspeni @ericw_ai Thank you lol yeah not sure why he did blurry.
English
@ericw_ai The above one-pager was actually published about 7 months ago, so it didn't just drop.
I don't understand why is is blurred, but here is the full report with all the information. www-cdn.anthropic.com/58284b19e702b4…
English
