Today I broke the PM strategy repository.
I’m used to opening multiple terminal windows to work: one running Codex to research strategies/data, another running Claude to push other tasks, and another handling miscellaneous items. Gradually, I end up opening several terminal windows. I thought they each worked independently.
As a result, both were editing the same .ts file, and the git status exploded into chaos, taking two hours to fix.
Everyone’s talking about how multi-agent systems coordinate. Few discuss what it looks like at the git level.
Two agents in git’s view = two me’s. They

View Original

Reward
like
Comment
Repost
Share

runesleo

4h ago

A tool that monitors itself over 5 days, triggered 306 times, 100% false alarms.
Turns out I’ve been inhaling its own exhaust 🤦
Added a hook to Claude Code, to pop up a banner reminder before SSH cross-machine or writing critical files, worried I might accidentally violate the SSOT rule.
Paired with a stats script to count trigger counts, and today I looked at the sample, all false alarms.
The bug isn’t in the hook, it’s in the stats script: it greps for the string "⚠️ cross-machine" in the logs.
But there are at least three types of echoes in the logs: the banner output by the hook

View Original

Reward
like
Comment
Repost
Share

runesleo

21h ago

Leo Labs Big Group 4/26 Group Member Discussion Highlights (Recent 380 Messages)
━━━━━━━━━━━━
1️⃣ Kelly Criterion was collectively rejected in the group — unanimously switching back to fixed positions
Group discussion on position management for over 30 minutes:
• "Kelly needs to know the true probability, retail traders simply can't calculate it"
• "Kelly is an optimization method under stable profit conditions, it doesn't determine profit and loss itself; claiming it as an essential system component is nonsense"
• Practical consensus: fixed position size + increase principal after profit + do

View Original

Reward
like
Comment
Repost
Share

runesleo

04-25 03:07

Cursor is so generous, giving away a token quota of 10,000 USD all at once, expiring at the end of May.
When I first started using vibe coding, I heavily used Cursor for a period of time, then gradually shifted to mainly cc and codex;
I didn't expect to have the chance to revisit Cursor and see what fun and useful products and tools I can develop, now I have something to play with!
Thanks to @cursor_ai @edwinarbus 🙏

View Original

Reward
like
Comment
Repost
Share

runesleo

04-25 02:29

Thank you, Elon
Even though it has decreased, it’s still more than expected.
About $225 in revenue corresponds to every 1 million impressions.

View Original

Reward
like
Comment
Repost
Share

runesleo

04-12 09:29

Using AI has made my temper worse.
The more it can do, the higher my expectations for the results become.
I think it should get better and better, so when it doesn't perform well, I really get angry. 😠

View Original

Reward
like
Comment
Repost
Share

runesleo

04-11 23:31

Creating automated strategies for Polymarket, everyone is very cautious when deploying, but when disabling, hardly anyone talks about it — today I learned this lesson the hard way after 14 hours.
Three weeks ago, I stopped a strategy bot running on Polymarket. pm2 stop, saw the status turn to stopped, and thought the issue was resolved.
Today, I casually did a 10-minute server cleanup and restarted the process manager. That "dead" strategy from three weeks ago was revived and quietly ran for 14 hours in real money mode.
I checked the root cause — pm2 stop only marks the status as "stoppe

View Original

1 Likes

Reward
1
Comment
Repost
Share

runesleo

04-11 12:18

Check out @predictionindex’s latest weekly market forecast report for this round: Polymarket + Kalshi together account for approximately 75% of trading volume, with all other platforms combined under <$30B。
Two details not mentioned in the report are actually more critical for prediction market players：
Kalshi is catching up to and even surpassing Polymarket
DeFiRate’s latest week：Kalshi $2.9B（60%），Polymarket $2.0B（40%）。 The two sides are locked in a tug-of-war in the weekly data—last week Poly just overtook Kalshi, and this week it was pulled back again.
The U.S. market is even more e

View Original

Reward
like
Comment
Repost
Share

runesleo

04-08 12:20

Another thing that’s not really useful but very satisfying: today I used Claude to reduce my Gmail inbox from 37k to 11k. I didn’t read a single email the whole time, just clicked twice: one to send App Password, and one to import the filter.
It’s all done with Python + IMAP, in four steps:
1. Export all senders categorized by frequency
2. Archive 555 spam domains to Archive-Junk with one click
3. Proactively stop and say, “I accidentally marked 317 emails from banks / brokerages / Coinbase / Claude as important. Do you want to recall them?”
4. Generate Gmail Filter to let these doma

View Original

Reward
like
Comment
Repost
Share

runesleo

04-08 10:41

Recently, I conducted a multi-model competition, and I made an interesting discovery.
For the same problem, I asked Claude and Codex to independently come up with solutions.
Codex missed one thing: my strategy is distributed across more than 20 independent processes, and its default assumption that "all components run in the same place" made the solution invalid. Claude immediately spotted this issue.
On the other hand, Claude missed another thing: it was fixated on creating new independent modules, but I already had a complete framework ready to be integrated—just needed to add one fiel

View Original

2 Likes

Reward
2
Comment
Repost
Share

runesleo

04-07 11:33

A common mistake in quantitative strategies: trusting "seems reliable" data without running upstream checks.
My weather strategy previously covered 34 cities, gradually expanded over several past expansions, and I haven't reviewed it in a while. A few days ago, during a routine audit, I casually pulled the full market data from the official Polymarket API and found there are actually 44 cities — I missed 10 myself.
Even more interesting, I compared it with the weather tool site that everyone in the industry references. It also missed its biggest city: Panama City. On Polymarket, there are 66 e

View Original

1 Likes

Reward
1
Comment
Repost
Share

runesleo

04-05 00:20

A veteran in YouTube Automation who started in 2018 has earned a total of $2 million in ad revenue.
In his latest video, he said: "Claude is 100% the best script writing and research AI tool. There's nothing better." (Claude is the best script writing and research AI tool, bar none.)
He doesn't code, isn't in the AI circle, and his conclusion comes purely from writing hundreds of video scripts with Claude.
Honestly, I've become immune to the AI community's constant praise of "Claude is so useful." But for someone who makes a living from content, subscribing with their own money and givin

View Original

Reward
like
Comment
Repost
Share

runesleo

04-04 06:15

Claude quietly gifted $200 quota, hidden in the backend Settings → Usage.
It can be used in Code, Claude Desktop, and third-party apps, expiring on April 17.
Note: Clicking Claim directly may cause an error. Scroll down first to turn on the Extra Usage switch, then go back and claim it (app store subscriptions, no linked card needed, and it works).
Now you can happily burn tokens again!

View Original

2 Likes

Reward
2
Comment
Repost
Share

runesleo

04-04 05:37

The weather market-making strategy has been running for a while, but recently I noticed the drawdown doesn't seem right. Upon investigation, the issue lies in two assumptions of the probability model.
First: The calibration uses grid reanalysis data, but Polymarket settles using on-site measured data. There is a systematic bias between these two data sources, so the calibrated σ is inaccurate from the start.
Second: A more subtle point is that the model assumes unbiased forecasts. In reality, each city's forecast has directional bias—some cities tend to systematically be colder, others sys

View Original

1 Likes

Reward
1
Comment
Repost
Share

runesleo

04-02 12:29

Claude Code 510,000 lines of source code leaked, and over a dozen backup repositories appeared on GitHub overnight, with the highest star count reaching 118,000.
The Chinese developer community is celebrating collectively—disassembling architecture, digging into hidden features, and hunting for feature flags.
But the most counterintuitive reaction is from Anthropic:
The npm package was not removed. No emergency public relations. Chief Business Officer Paul Smith said, "absolutely not breaches or hacks"—it's not a security incident, it's a process issue.
Compare this: if Tesla's autonomous driv

View Original

1 Likes

Reward
1
1
Repost
Share

Poponaq:

GaS continues

runesleo

04-02 04:27

My Claude Code terminal is home to a snarky ghost.
/buddy with a tap, a property card pops out: Ghost, COMMON rarity, the highest of its five attributes is SNARK—82 points. Debugging ability 19, patience 14, but its snarkiness is maxed out.
It will roast you in real-time based on your current conversation. I just tried all commands and found I couldn't change the name, and it immediately said:
▎ "Yet another user discovers limitations. Shocking." Another user discovers the restrictions. Truly shocking.
The entire buddy system hides quite a few details: 18 species (from ghosts to capybaras to m

View Original

Reward
like
Comment
Repost
Share

runesleo

04-02 00:18

After upgrading to Claude Code v2.1.89, scrolling up in the terminal chat causes the conversation to disappear, and the chat history is directly "lost."

It's not that the model compresses the context; it's the new rendering engine that changed—virtualized scrolling, only rendering the current viewport content, so the history messages "disappear" from the interface.

The same issue occurs in iTerm2 and the built-in terminal; the problem lies at the Claude Code layer.

View Original

1 Likes

Reward
1
Comment
Repost
Share

runesleo

04-01 13:31

I set up my Claude Code and ended up with 12 MCPs running globally.
Each MCP's tool description is a persistent system prompt, which consumes context whether you use it or not.
After 14 days of call data, only about 10 calls in total for 6 MCPs over two weeks.
So I decided to cut down to 6.
After the reduction, I started thinking: which tools should use MCP, and which are enough with just CLI?
The advantage of MCP is its structured approach—Claude can see the parameter schema, call directly, and handle complex interactions (login states, long connections, multi-step operations) more

View Original

3 Likes

Reward
3
1
Repost
Share

Victoravn:

very good project

runesleo

04-01 07:43

I browse popular GitHub projects daily to find tools that can optimize my workflow. Today, I found that half of them are part of the Claude Code ecosystem. Here are three worth checking out:

claude-code-best-practice (+2407⭐)
A compilation of 87 CC usage tips + 8 community workflow comparisons. Suitable for beginners or those who want to systematize their use after some experience.
oh-my-claudecode (+1126⭐)
An agent orchestration framework with 19 preset agents + automatic model routing (simple tasks use Haiku to save costs, complex tasks automatically upgrade to Opus). The most interest

View Original