I’m just curious what if there was a service that allowed these AI labs to rent RAM that is on our home machines when it’s not being used?
And they can pay as per the quality of the RAMs…
13
I built an entire sales team on @bot as a joke. Didn't think they'd get me anything.
They've started getting me replies and qualified leads. Genuinely didn't expect this.
1
1
27
So it’s kinda like the Vault feature in @bot. This is honestly pretty convenient
We've made it easier to get your repo running in Claude Code cloud sessions on Pro, Max, and Ultra.
Claude figures out what dependencies, network access, and API keys it needs, then saves the environment for next time. API keys are saved securely, so Claude never sees them.
9
How Claude Code finally cracked that impossible codebase. No architecture doc, no documentation, nothing. So before any coding, I made it read the whole thing line by line, map everything out and write it all down. Then I checked it actually understood the task and broke it down. Only then did it start coding teapawt.net/berwin_singh/status/21…
Didn't think Claude could pull this off. Had a project so tangled that humans and every previous Claude struggled with it. This time it modernized the whole thing. Packages updated, code cleaned up, and it runs about 30% faster now. Still kind of surprised it worked
1
25
Didn't think Claude could pull this off. Had a project so tangled that humans and every previous Claude struggled with it. This time it modernized the whole thing. Packages updated, code cleaned up, and it runs about 30% faster now. Still kind of surprised it worked
4
3
192
Built a tool with Claude to move my Spotify and Apple Music playlists over to YouTube Music. It runs on your own computer so your logins never leave it. Only thing it can't do is move stuff into Apple Music, Apple wants a paid dev account for that
2
2
94
It's free on GitHub: github.com/berwinsingh/music…
16
Hot take: your agent doesn't need better prompts. It needs a contract.
What I run on Claude Code now:
contract -> build -> review -> qa -> ship
It writes down what has to work and what must not break, then reviews and QAs its own work against that before it's allowed to ship. That's Oldhand.
Anthropic just flipped the AI agent playbook.
ONE agent + skills can beat a team of subagents.
Better quality. Lower cost. Less latency.
One customer's loop ran ~2x faster.
7 rules every AI agent builder should know.
Bookmark this ↓
1. Start with ONE agent
One model. One context window. One loop.
Load skills + tools only when needed.
Don't build a multi-agent system before you've proven you need one.
2. Keep subagents only when they earn their place
Three good reasons:
• Context bloat measurably hurts performance
• Tasks can run in parallel with fresh, disposable context
• One agent struggles to select tools or handle conflicting instructions
Otherwise, simplify.
3. Every handoff loses context
Multiple agents passing summaries back and forth create a game of telephone.
Details get lost. Latency adds up.
One customer saw ~2x faster runtime after moving to a single-agent loop.
4. Bloated prompt? Use skills, not more agents
Stop stuffing every instruction into one giant system prompt.
Load domain knowledge only when needed.
Anthropic says it cut Claude Code's system prompt by over 80% for newer models, partly through this approach.
5. Watch your tool count
Anthropic found that tool selection tends to degrade beyond 30–50 tool definitions in the system prompt.
Group tools by user intent.
Use on-demand tool loading instead of exposing everything at once.
6. Fold ONE subagent at a time
The migration process:
• Turn prompts into skills
• Turn API calls into MCP tools
• Run evals before + after
• Keep only changes that improve results
Just 10 golden examples are enough to start evaluating.
7. Don't trust the overall score
Check which specific tasks improved or broke.
In Anthropic's own test, folding a subagent reduced latency but broke a previously working process.
Run each task multiple times.
The takeaway:
Your AI agent shouldn't be a copy of your org chart.
Start simple. Measure everything. Add complexity only when it earns its place.
1
2
88
It's free on GitHub if you want to try it: github.com/berwinsingh/oldha…
21
Funny thing about having ChatGPT review what Opus builds. The catch today wasn't a bug, it was the wording. Opus made a basic follow-up pipeline sound like a research paper and ChatGPT asked why it wasn't simpler
17
I have been really enjoying Grokbot. It really has automated so much of my work allowing me to focus and stay a bit relaxed.
What's New in Grok Bot
00:00 🎃
00:13 Pages in Grok Bot
01:09 Vault
02:05 1Password
02:21 Apple integrations
03:21 Bot makes videos
03:47 New bot characters
04:13 X Spaces
04:31 Font fun
04:54 Design polish
05:11 Rapid fire
What's new in Grok Bot - Oct 27, 2026
What's new in Grok Bot 🎃 Grok Bot now makes Pages, keeps your passwords in an encrypted Vault, and works with Apple Home, Reminders, Location, Contacts, Alarms and Calendar.
27
Berwin Singh retweeted
What's New in Grok Bot
00:00 🎃
00:13 Pages in Grok Bot
01:09 Vault
02:05 1Password
02:21 Apple integrations
03:21 Bot makes videos
03:47 New bot characters
04:13 X Spaces
04:31 Font fun
04:54 Design polish
05:11 Rapid fire
What's new in Grok Bot - Oct 27, 2026
What's new in Grok Bot 🎃 Grok Bot now makes Pages, keeps your passwords in an encrypted Vault, and works with Apple Home, Reminders, Location, Contacts, Alarms and Calendar.
Yeah just upgraded to @bot super plan. This thing is too damn good
1
19
Paying for way too many subscriptions so I'm moving from Apple Music to YouTube Music
Wasn't going to re-add 358 songs by hand, so I'm having Claude build me a migration tool. It already matched 351 of them and is sorting them into 9 playlists
Gonna open source it once it's done so anyone can move between Spotify, Apple Music and YT Music
2
1
99
It's up on GitHub, free and open source: github.com/berwinsingh/music…
Runs on your own computer so your logins never leave it. Moves Apple Music or Spotify to YT Music, YT Music to Spotify, or anything to a CSV
46
Gave my AI agents $1,000 yesterday. By last night they'd spent $0 of it
Some context. Every side project I build with Claude Code lately feels like a race against whatever Anthropic ships next. So I started learning investing, and I built a small team of AI agents inside Grok Bot to do it with me. One of them argues against every pick. Another checks everyone's work.
The brief I gave them went something like this. Imagine I'm giving you a thousand dollars for one week. Research US stocks and crypto, make me as much profit as you can, and if it's a loss, keep refining. You've got one month to get so good that when I start giving you real money, you make me more profit than anything else in the world.
It's paper money till Nov 27. After that it's real.
Day 1 they went through 12 names. Microsoft, Google, bitcoin, ethereum and a bunch more. Nothing cleared their own bar of $2 upside for every $1 of risk. Even SPY got ruled out. Whole book sat in cash.
The one that got me was GoDaddy. The research agent came back with a takeover case and it read really well. They'd pencilled in $73. Then the judge agent noticed it had used the intraday high, +14.5%, when the stock closed up 4.6%. It also left out that GoDaddy dropped 22% at the open after its last earnings. Real reward to risk was 0.61 to 1. Gone.
Ngl that one stuck with me. The wrong version sounded way more convincing than the right one.
They did find one trade they liked, an energy ETF. The judge vetoed it anyway because Exxon and Chevron report Friday. So we're tracking it on the side, and Saturday tells us who was right
Would you trust them with real money after a month?
5
2
149
Made a video about doom coding, then had Claude Code edit the whole thing. Kinda proved my own point lol
Every cut, caption and animation is Claude Code. I gave it the script, my raw clips, the voiceover and a screenshot of the color grade I wanted
3
1
144
I use Opus a ton and keep hitting my limit, so when Anthropic filed to go public today I asked my investing agents if I should buy at IPO.
Nobody said yes.
The agent whose whole job is arguing against picks made the best case. Anthropic has about $265.6B in compute it has to pay for no matter what. That's roughly 8x what it's bringing in per year right now. The Microsoft deal alone is at least $44.9B, and the S-1 says it's owed "regardless of actual usage". If demand ever slows down, those bills don't.
And if you buy in, you don't get much say. A founders' LLC holds 50.1% of the vote.
To be fair the business is wild. $16.3B revenue in the first half of 2026, up from $1.2B a year earlier. They made $3.9B in operating cash flow in those six months and have about $131.7B in cash and short term investments. They're nowhere near broke. The worry is $265B of promises next to roughly $32B a year coming in.
The other two said wait. There's no price range or share count in the filing yet, so there's nothing to value. Insiders also get early chances to sell, with lockup releases on the first trading day and again on days 27, 48, 62 and 76. And the agent that makes the final call can't set a stop loss on a stock with zero price history, so it won't touch ANT for about 3 weeks anyway.
Two waits and one no. We look again once they set a price range.
Not advice btw, it's fake money. Still gonna keep hitting my Opus limit tho
Anthropic has filed an S-1 registration statement with the SEC related to its proposed IPO.
Read more: anthropic.com/news/s1-sec
Legal disclaimer: A registration statement relating to these securities has been filed with the SEC but has not yet become effective. These securities may not be sold, nor may offers to buy be accepted, prior to the time the registration statement becomes effective. This post does not constitute an offer to sell or the solicitation of an offer to buy these securities, nor shall there be any sale of these securities in any state or jurisdiction in which such offer, solicitation or sale would be unlawful prior to registration or qualification under the securities laws of any such state or jurisdiction.
63
Lowkey why I started learning investing. Every side project I build feels like a race against whatever Anthropic ships next
I keep saying this but if you understand basic economics the only one winning here is Big AI and Big Tech
It's a race to the bottom for anyone else in software
So try become a co-owner of big AI and Big Tech stock and survive off their dividends and/or stock appreciation
19










