AI Agents • Automation • Open Source AI Tools, insights & experiments. ⚡ DM for collaborations.
Chicago, IL
Joined February 2020
- Tweets17.8K
- Following38.4K
- Followers47.1K
- Likes41.8K
Pinned Tweet
I’ve been exploring Zoqo and learning more about how prediction markets actually work.
One interesting thing I’ve learned is that market prices aren’t just numbers. They represent implied probabilities based on what traders are willing to pay for an outcome.
Being part of the Zoqo Build-Off gives me an opportunity to go beyond reading about these systems.
I’m testing a trading bot, observing its decisions, learning how the markets behave, and understanding the risks involved.
14 days. 25 creators. One bot each.
I’ll be sharing my experience, lessons, and discoveries along the way.
Let’s see what I learn. 👀
@ZoqoTrade
🤝 Paid partnership
Almusty Labs retweeted
Automating a messy workflow just makes the mess happen faster.
Most of the time the real problem isn’t speed - it’s the duplicated steps, the extra approvals, and the parts that keep failing. Mapping those first usually shows that half the process doesn’t need to be automated at all.
What’s a workflow you tried to automate that turned out to be broken underneath?
18
15
442
Most traders don’t need another dashboard they have to check all day.
They need fewer, better alerts.
An agent can watch price, volume, and news in the background and only surface the things that match conditions you actually care about. The hard part isn’t monitoring - it’s making sure the alerts are specific enough that you don’t start ignoring them.
What’s an alert you currently get that usually isn’t worth your time?
4
11
192
Almusty Labs retweeted
I’ve been exploring Zoqo and learning more about how prediction markets actually work.
One interesting thing I’ve learned is that market prices aren’t just numbers. They represent implied probabilities based on what traders are willing to pay for an outcome.
Being part of the Zoqo Build-Off gives me an opportunity to go beyond reading about these systems.
I’m testing a trading bot, observing its decisions, learning how the markets behave, and understanding the risks involved.
14 days. 25 creators. One bot each.
I’ll be sharing my experience, lessons, and discoveries along the way.
Let’s see what I learn. 👀
@ZoqoTrade
🤝 Paid partnership
The research workflows that are easiest to fix are the ones where every stage has a clear input and a clear expected output.
When something goes wrong, you can usually point at the exact step that broke - the sources, the extraction, the summary - instead of just knowing the final result looks off.
Most of the time the problem isn’t the model. It’s that one stage was never defined well enough to test.
Where does your current research process usually break?
9
12
330
Almusty Labs retweeted
Thanks to the verified accounts who followed me this week! Give them a follow:
@AlmustyLabs AI agents and automation tools
@JeremyCMorgan Engineer and instructor helping devs learn AI
@Sadiapv Exploring life every day
So glad you're here. Stop by and say hi!
ALT AI image created by @buckleyplanet
Automating a messy workflow just makes the mess happen faster.
Most of the time the real problem isn’t speed - it’s the duplicated steps, the extra approvals, and the parts that keep failing. Mapping those first usually shows that half the process doesn’t need to be automated at all.
What’s a workflow you tried to automate that turned out to be broken underneath?
18
15
442
Almusty Labs retweeted
Polymarket traders, how much time do you spend watching markets, tracking price movements and waiting for the right moment to enter or exit a position?
What if you could automate some of that work? That's where Zoqo Trade comes in.
Zoqo Trade is a no-code trading automation platform built for Polymarket. It allows you to turn your trading strategies into automated bots without writing a single line of code.
Here's what you can do with it:
• No-code automation: Create custom trading bots based on your preferred market conditions.
• Custom strategies: Set entry triggers, take-profit and stop-loss levels, and define your preferred position sizes.
• 24/7 execution: Let your bots monitor markets and execute trades based on your predefined conditions, even when you're away.
• Risk management: Set strategy budgets, daily loss limits and cooldown periods to establish boundaries for your trades.
• Telegram notifications: Receive updates on order executions, settlements and other automation activities.
• Trade transparency: Review your trading history and understand why your bot entered or skipped a trade.
The interesting part is that you don't need programming skills or have to spend your entire day staring at charts to put your trading strategy into action. You define the rules and @ZoqoTrade handles the automation.
But I'm not just here to tell you about Zoqo. I'm here to explore it myself. I'll be sharing my experiences as I use the platform, the strategies I test, the profits, the losses, and everything I learn along the way.
If you're a Polymarket trader, you might want to stick around. There's more coming.
Explore more about Zoqo Trade: zoqo.trade/
Almusty Labs retweeted
The tool that usually moves the needle most isn’t the one that sounds most impressive.
For a lot of agents it’s just reliable web search or the ability to actually run code. Browser control and API access are useful later, but only if the core task is clear.
What’s the one tool that would change how useful your current agent is?
Almusty Labs retweeted
One lion vs an entire army 🦁⚔️
This is my entry for the OiiOii × Kling 4.0 Creative Contest. The theme is "The Journey," so I made "The Last King."
A lone lion charges through a wall of shields, gets surrounded by an entire army, and roars his way out, all to make it home to his family by sunset.
Generated with Kling 4.0 on @OiiOii_AI.
There's a $5,000 prize pool, with $2,000 for first place. Entries close Nov 22. Come make your own journey and join me 👇
Join Now - oiioii.ai/campaigns/dd5046a0…
@Kling_ai #OiiOii #Kling4 #OiiOii_Kling4CreativeContest
🤝 Paid partnership
🎬 OiiOii × Kling 4.0 Creative Contest is HERE! 🌍
Every journey tells a story. You call the shots.
This time, the theme is “The Journey”. Explore diverse worlds with OiiOii × Kling 4.0, bring every scene to life, and turn your imagination into a cinematic adventure!
🏆 Win up to $2,000 in prizes!
Plus, 9 additional awards across 3 creative categories:
📖 Story & Narrative
🎥 Cinematography
🎨 Visuals & Style
📅 Contest Period: November 9–22
Almusty Labs retweeted
More time to work on your projects without paying for another annual subscription.
That's what Flova's Black Friday offer brings, alongside 40% off Seedance 2.5 for 365 days.
flova.ai/?refCode=U3FQ8YHE
Almusty Labs retweeted
What if your learning journey was shaped by your goals rather than a one-size-fits-all university curriculum? 🎓
That's the vision behind A School of One from Paradigm Study.
Want to learn Spanish, develop AI skills, or tackle a challenging subject? Imagine having a personalized course designed around your needs and learning goals.
With on-demand courses, an infinite learning canvas, and Clover offering real-time explanations, hints, and feedback, learning could become more flexible and interactive.
But there's one question I find particularly interesting:
If AI-native schools introduce certificates in the future, will employers recognize them? What would you look for first — practical skills, real-world projects, or independent assessments?
The future of education isn't just about gaining knowledge. It's about proving what you can actually do.
Explore: paradigm.study
@ParadigmStudy
🤝 Paid partnership
Almusty Labs retweeted
An agent that can read your email and access your files doesn’t also need permission to move money.
Most people give broad access because it’s easier to set up. Then they spend time cleaning up actions the agent wasn’t supposed to take.
The safer default is the opposite: only the tools the task actually requires.
What’s the most unnecessary permission you’ve seen an agent given?
Almusty Labs retweeted
Can AI help a team explain an idea to each other? A rough draft gives the director something to say “no, not that” to before a costly setup is built. Sometimes the hard part is getting the idea out of one person's head.
🤝 Paid partnership
What changes when filmmakers can see a scene before they shoot it? 🎬
Meet Next Frame, BytePlus' original video series.
In this episode, Young Washington director @JonErwin13 shares how he brings AI into filmmaking, from developing a script to creating visual effects on set.
Almusty Labs retweeted
Giving an agent every tool it might need is usually how you end up with unexpected actions and permission issues.
Most of the time it only needs the few tools required for the actual task. Everything else just increases the chance it picks the wrong one or does something you didn’t intend.
Have you ever had to strip tools back after an agent started doing things you didn’t want it to?
The tool that usually moves the needle most isn’t the one that sounds most impressive.
For a lot of agents it’s just reliable web search or the ability to actually run code. Browser control and API access are useful later, but only if the core task is clear.
What’s the one tool that would change how useful your current agent is?
Almusty Labs retweeted
What if your trading strategy could make decisions even when you're not watching the market?
I created my first automation on @ZoqoTrade using the Final Favourite 5m template.
Here's how it works: The bot watches a 5-minute BTC market and buys whichever side is leading in the final 30 seconds.
No need to sit in front of your screen waiting for the right moment. The bot follows the rules automatically.
I'm putting my automation skills to the test in the Zoqo Trading Competition, where traders can create bots to trade for them.
Explore: app.zoqo.trade/
Almusty Labs retweeted
I just started running an automated trading bot on @ZoqoTrade, and I've been spending quite a bit of time figuring out how to get the setup right.
I've always been curious about automated trading, so I decided to use this competition to build a strategy, test it and see how it performs in live markets.
Here's how I set it up, what I tested and the results so far.
🧵⤵️
Almusty Labs retweeted
I ranked 8 data centre GPUs for serving inference by what one rental dollar buys in memory and memory bandwidth.
When a model generates tokens, memory decides which models fit on the card and bandwidth sets how fast tokens come out. This leaves out compute and interconnect, which matter more for long prompts and multi-GPU serving.
S
> MI325X: 256 GB at 6 TB/s for $3.10 an hour. The cheapest memory and the cheapest bandwidth of any card here.
> MI300X: 192 GB at 5.3 TB/s for $2.86. Almost the same value, one generation older.
A
> B300: 288 GB, the most memory on a single card. Large MoE models need fewer GPUs per copy.
> H200: 141 GB at 4.8 TB/s for $5.00. The same bandwidth per dollar as the B200 at about 60% of its hourly price.
> B200: 7.7 TB/s for $8.01, the second-fastest memory on the list.
B
> RTX PRO 6000: 96 GB for $2.38. Cheap capacity, but at 1.6 TB/s it's slow on large models.
> H100: $3.94 an hour for 80 GB. The most expensive memory per GB on the list.
C
> L40S: 48 GB at 0.86 TB/s. The lowest hourly price and the highest cost per unit of bandwidth.
The two best-value cards are listed by 3 to 5 providers each. The H100 is listed by 45. Teams whose serving stack runs on AMD are paying the least per gigabyte today.
