@adolandev
Account based inWest Asia
About this account
- Account based in
- West Asia
- Connected via
- West Asia App Store
Account-level information from X, not a live location or the device used for a specific post.
I build things. I break things. Occasionally on purpose. Ambassador @cognition.
Joined June 2026
- Tweets853
- Following353
- Followers398
- Likes3.9K
Pinned Tweet
Resetwatch is live. Remaining quota and reset clocks for @NousResearch Hermes Desktop.
How full each window is, how much is still there, and when it comes back. Nous, Claude, Codex, Cursor, Kimi, with the plan name on the card. No chat has to be open.
github.com/Adolanium/hermes-…
I can see why API dollars are a bad measure of a subscription. But “more work done” needs a before-and-after on the same workload. Otherwise it’s hard to know what $200 buys you.
Hi,
Tomorrow we are re-opening the Pro $200 subscriptions to new subscribers, but together with it we are also changing how we calculate the usage for it. In effect, if you do the math, it will net out at half the dollar in API spend compared to the old Pro $200 plan.
Now that it's said, let me explain why this is happening and why you will still get more work done than if you were on the Pro $200 subscription one month ago.
(a) We didn't want to compromise in other ways and are committing to not reintroducing the 5h limit, so that you can fully use the weekly usage when you want.
(b) On the subscription, we guarantee that over time you always get more work done and with an increasing level of quality. This means that you will continue to get more value per dollar spent as a result of models getting more efficient and us passing down the improvements in the form of API price reductions.
(c) We don't want to put an incentive on ourselves to artificially inflate the API list prices to make it look like you are getting a lot (and workaround it through discounts, etc). Instead we want to continue to both rapidly reduce prices and increase capabilities of models on the API. This week we introduced GPT-6 Sol and GPT-6 Luna at 50% of their previous price. Over time, we see prices go low enough that it makes sense for most to buy usage as needed without there being a significant gap between what you get in a subscription and what you get in the API for a dollar spent.
(d) Tomorrow, we are adding more things to the subscription that won't draw on the usage, I won't reveal what that is yet.
I wanted to be transparent before all the big announcements tomorrow. Lots of new exciting things are coming to the subscriptions that will make it super compelling, but I wanted to make sure to share this change ahead of time so you can all understand it before we shower you with good news.
Codexingly,
Tibo
Flight 14 and watching a rocket this big fly still hasn’t gotten old.
Starship Flight 14 launches in ~30 mins
nitter.cf/i/broadcasts/1qxvvenMP…
Adolan retweeted
Hermes Desktop, in any browser 🌆
My @NousResearch updated PR. `hermes webapp` serves the real Desktop app from your Hermes host: chat, files, Git, a terminal that survives a refresh, and sign-in for remote binds.
Look forward to it.
github.com/NousResearch/herm…
I’m out of the house with only a Windows machine today. My MacBook is at home, and I wanted to work on an iOS app.
A while ago, that would have been the end of it. Today I gave Devin a task, and now I’m watching it use a Mac and an iOS Simulator through a live stream in my browser. I can see the app running, see what Devin is doing, and step in if I need to.
The part that gets me is that the simulator is actually useful to the agent. It can inspect the app on screen, work on the UI, and check what its changes look like in the same environment. I don’t need to have Xcode installed on the machine in front of me to follow along.
This picture is my Windows PC, away from home, with Devin working inside an iPhone simulator on a Mac somewhere else. Really impressive work from @cognition and @DevinAI.
This is a much bigger deal than just “watching the agent work”.
The agent gets its own remote desktop and persistent browser session. You can take over for logins, 2FA, CAPTCHAs, or anything that needs human input, then hand control back and it continues where you left off.
This is what human-agent handoff should feel like.
Really love where Hermes Desktop is heading. Congrats to @Teknium and the team!
You can now watch your agents work live as Hermes Desktop streams the screen of any bot's screen or session in real time.
Watch it drive a browser, type into a terminal, or open a window. Take over anytime to interact with the Bot Screen or type credentials and seamlessly hand back off when you're done.
Luna is the most interesting point here. Almost the same score as Opus at roughly 1/22 the cost. Makes me wonder how far you could get using Luna first and escalating only the tasks it can't finish.
Agreed. You don’t have to win every benchmark to offer a compelling tradeoff.
On mobile, I’d want to share a page to the agent, give it a task, and let it work in the background. Notify me when it needs a decision, let me take over without losing progress, and show a clear receipt of what it actually did.
What features do you need in a Browser Use app?
Share your answers in a repost:
First 20 people get early access on TestFlight (only iOS) and $50 for the agent to spend.
The best part of the Hermes plugin catalog is finding something you didn’t know your agent could do.
You get to see what other people find useful, install it, and try it in your own workflow.
Great walkthrough from @tonbistudio, featuring two of my plugins!
Hermes Agent now has a built-in Plugins Catalog!
I made a short video introducing the catalog and how to quickly install plugins, and tried out three of them:
- resetwatch by @adolandev
- hermes-newswire by @tonysimons_
- hermes-office by @adolandev
Let me know your favorite plugin and I'll check it out!
Keep the core lean, make plugins easy to find, and let people add what they need. Less to configure upfront, less for the core team to maintain. Definitely a step in the right direction for Hermes.
"AI is the future" is becoming the justification for paying any price in the present. That's how bubbles work.
Instinct is in talks to raise about $1 billion at a roughly $10 billion valuation, less than a month after raising $250 million at a $2.25 billion pre-money valuation.
The 23-year-old founder Noah Shinn is building a personal AI agent that can operate software and complete tasks like answering emails, negotiating bills and booking reservations.
Instinct now has more than 100,000 users and has already run into compute capacity constraints as demand grows.
Its valuation has gone from roughly $50 million in the spring to $10 billion under discussion today.
Source: The Information
An agent edits a file, then changes it back.
The final diff is empty. But those edits still happened.
Session Diff for @NousResearch Hermes Desktop lets you inspect a run’s file changes, including the ones it undid.
Click a file. See what happened. No digging through the tool log.
github.com/Adolanium/hermes-…
I've built 8+ plugins for Hermes, and they're now easier to find.
Grab them from the new plugin catalog, right inside Hermes Desktop.
What should I build next?
Toolsmith is live. Everyday developer tools inside @NousResearch Hermes Desktop.
Format JSON, decode Base64, inspect a JWT, compare text, generate hashes. 19 tools, one sidebar entry, all processed locally. No API keys or model calls.
Chain steps into recipes: URL decode → Base64 decode → JSON format. See what changed at each step, then save the recipe for next time.
Need the result in a conversation? Review it and add it to your chat draft.
The little tools you keep opening browser tabs for, now inside Hermes.
github.com/Adolanium/hermes-…
Still kinda crazy seeing this written out.
1,393 subagents.
19 hours.
A million-line Python codebase.
34.4% smaller by the end.
This is exactly the kind of ridiculous experiment that makes working around Hermes so much fun.
We’re living in a very weird and exciting era of software engineering.
New blog post:
We had a million lines of Python to clean up. On September 2nd @Teknium asked Hermes Agent to do it.
1,393 subagents and nineteen hours later, the codebase was 34.4% smaller, saving us nearly $2m in engineering hours.
nousresearch.com/refactoring…
Wait, this might actually be a big deal.
A frontier model that cannot generate text.
Instead, it’s built purely for decisions:
- 20-200x faster
- $0.042 / MTok input
- free output tokens
- cheap enough to call continuously in real-time software
Maybe forcing every AI model to "speak" was never the endgame.
After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI?
I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev
• 20-200x faster
• 40-400x cheaper (w/ output tokens free)
• Frontier composable intelligence optimized for decisions
AFAICT the shortest path to AI-based economic revolution
This makes me more interested in trying 4.8. Being specific about what still needs fixing gives me more confidence in the update.
I'm enjoying @devindesktop . However, on the $20 plan, a day of using Fusion with Fable 5.1 Medium + SWE 2.0 High used half my weekly quota. Anyone on the $200 plan using the same setup? How far does the quota go for you?
