Building @whirl_sh — AI orchestration layer for operations. ex-@monzo, @Citymapper
London, England
Joined April 2011
- Tweets2K
- Following804
- Followers1.1K
- Likes2K
Pinned Tweet
I'm opening @whirl_sh for early access
Most AI agents still sit idle, waiting for a prompt. Not good enough for ops teams & messy real-world problems
We need AI that runs in the background 24/7, proactively solving issues as they happen — with approval and visibility so it doesn't go fully rogue (🦞)
At @packfleet I spent four years watching our ops team keep a dozen plates spinning across spreadsheets, Slack, email, WhatsApp, phone calls, and other tools that didn't talk to each other
Whirl sits across your stack and proactively solves issues as they come in, leaving the team for the work that needs a human in the room
(my good friends at Idylliq are already running their Paris property ops on it. Guest comms, booking admin, cleaners, owners, maintenance...)
If your days are full of coordination work across a dozen tools, good chance Whirl can take a chunk off your plate.
DMs open for early access or sign up on the site
I thought I'd be typing less and less when coding with agents, but I'm typing way more. It makes sense though, IDEs, tab completion and copy/paste was the majority of writing code.
We might be about to split people into two: those who talk to their agents and forget how to type, and the rest of us learning to push 200wpm
3
166
There's just *no way* its design taste is better than Opus 5.5. Either OpenAI pulled a miracle in a point release or you guys have no design taste
My thoughts on GPT-6.2 Sol so far, based on my own tests:
It's a massive leap over GPT-6.1 Sol. What's even more impressive is that I'm getting better results than GPT-6 Astra, even in Max mode. I prefer its design taste to Opus 5.5's, and it's incredibly fast.
A few things really stand out:
- I've run dozens of Three.js, WebGPU, and HTML tests and barely used 10% of my $200 Pro allowance.
- It's creative from the first response. I don't have to keep asking for changes to get something beautiful.
- So many first attempts already make me go, “Holy shit, this looks incredible!”
The limits are honestly what impressed me most. Being able to keep trying ideas without worrying about suddenly running out makes a huge difference when you're using something every day.
And the best part? It's coming to Plus with generous usage limits too.
This is what I've wanted from OpenAI. Not just better answers, but enough room to actually enjoy using it.
1
117
Anyone here use @AmpCode? How do you do code reviews?
I can't seem to easily get Codex to review Claude's work and vice versa without creating multiple threads per feature . What's the best way to go about this?
(...i'm still trying to find a @conductor_build replacement)
3
1
154
The only thing that seems certain to me is that all jobs in the future will collapse to the role of a DRI (directly responsible individual).
Everything worth doing will be done with AI 100x smarter and faster, but it will never be responsible for things and can't be held accountable if something goes wrong. We essentially need someone to blame (or congratulate)
Good news is DRIs are already the most valuable and highly paid roles. Doesn't matter if you manage people or are a technical expert: the higher you get the more your role collapses to this
Bad news is I don't think everyone can do this. Maybe unemployment simply comes from not having enough people who want to be or are capable of being responsible
110
"All UI will be replaced by chat"
If you think this you clearly haven't seen how bad the average person is at putting their thoughts into words
3
111
Imagine you're stuck behind a slow truck, and FSD is like "hey you're running 10 mins late, want me to jump over that?"
119
It's been a while since I've heard anything about how well latest LLMs can play chess... So I made a quick app last night to see if I could beat the latest models: llm-chess-lac.vercel.app
TL;DR: they aren't that bad anymore! Still terrible by computer chess standards but they can actually play, and way above beginner level.
Argon managed beat me a couple of times (seems to be the best model for this), though Fable / Astra are still very beatable
Oh and you can chat / trash talk while you play
88
All inbound creates work for someone in a company and it's only gonna get worse from here. We can't have siloed customer service tools that live separately to the rest of the company, managed by a person.
We're building Whirl (whirl.sh) for that future. It's the point of contact of your company with the outside world.
We're also integrating personal agent protocol into it from the get go
2027 will be the year of the personal agent. It has finally found PMF with Instinct, Muse & Grokbot.
This will cause a tsunami of customer enquiries — it’s now essentially free for a customer to send hundreds of emails, texts and phonecalls to make a single purchase.
Businesses will need AI customer service tools to save themselves from drowning.
2
122
Robin Bilgil retweeted
Now live on the Chrome extension store! chromewebstore.google.com/de…
Also happy to announce that the team at @GPTZeroAI are kindly offering their AI detection API for free through Slop detector! Install & you get 1,000 free checks a week
Jev inspired this project and is great for classifying based on a prompt - just not good at detecting actually AI generated content. With GPTZero you now have the option for both
I've also open sourced the project here:
github.com/rbilgil/slop-dete…
Now live on the Chrome extension store! chromewebstore.google.com/de…
Also happy to announce that the team at @GPTZeroAI are kindly offering their AI detection API for free through Slop detector! Install & you get 1,000 free checks a week
Jev inspired this project and is great for classifying based on a prompt - just not good at detecting actually AI generated content. With GPTZero you now have the option for both
I've also open sourced the project here:
github.com/rbilgil/slop-dete…
Just wanted to check out Mistal Le Chonk. Fittingly they're the only AI lab with a full screen cookie banner
105
I've submitted this to the Chrome store for verification, with a little extra gift inside as a thank you for waiting. Link to be shared soon!
2
196
Maybe controversial opinion but I don't want a realistic AI avatar. It's always going to be icky uncanny valley teritory. Why dont we make them like cartoons or something, I wouldn't mind talking to a cartoon!
1
1
239
Looking forward to trying this but Google seems to be great at making models that benchmark well and are horrendous to use, so we'll see.
Introducing Gemini 4 Argon – our new frontier model.
It’s built for complex workflows across coding, enterprise knowledge work, and cybersecurity defense – rolling out today to a set of trusted testers through our Fairwind Program.
141
Hey Instinct it's my birthday and also my wife's birthday and also it so happens to be our anniversary tomorrow, isn't it extraordinary? Please find us a reservation ASAP
Instinct CEO Noah Shinn says agents will let restaurants give tables to birthdays and anniversaries over whoever clicks first
"And traditionally, because of the way that the internet has worked, at least for reservations, it's been first come first serve. So anyone that comes in, no matter how important or how not important it might be, whoever gets in first gets the reservation."
"But what if you had the ability for the agent to be able to communicate on both sides, communicate, hey, this is important. This is actually the person's spouse's birthday, the 30th birthday, a very special event is coming up."
"And the restaurant too can say, let's actually prioritize birthdays or major decade birthdays or major anniversaries or these certain people, I don't know, whatever it might be, right?"
"So now with the ability, with all of the logistical burden of having to describe exactly what your case is and what's happening and why it's important. Now we have the ability to perfectly match what does the restaurant want? What does the user want?"
"And that will result in more special events for the users being able to actually have a spot in the restaurant."
3
2
517