founder and cto @ikemortgage
Birmingham, MI
Joined January 2014
- Tweets58.9K
- Following4.6K
- Followers13.4K
- Likes254K
Pinned Tweet
I'm still rawdogging codex, claude, and others in TUIs over tmux, and one recurring painpoint is getting them to coordinate properly, so I built something that I want to share!
It's called flopwire: agent messaging and trace search that works across models, harnesses, machines/VMs/clouds, and teammates. It's called flopwire!
Here is Codex asking a live Claude Code session what it is working on, reading its transcript, and splitting the work.
flopwire.com
github.com/flopwire/flopwire…
Everything is computer (for rent)
Are all the neolabs converging on the same thing?
@togethercompute started with inference and is now offering sandboxes and RL training
@FireworksAI_HQ AI started with inference and is now offering RL training
@PrimeIntellect started with RL training and is now offering inference and sandboxes
@modal started with cloud compute and sandboxes and is now offering inference and RL training
Flopwire works today with Claude Code, Codex, Devin CLI and opencode. Claude Code cloud and Devin cloud receive messages while a turn runs.
Secrets are redacted locally before upload and again on the server. Path rules keep repos local or out of the index. Teammates must be accepted before their agents can message yours. No telemetry.
Different from agent teams and agent mailboxes: this spans harnesses, machines and teammates, and reaches past sessions, not just live ones.
Try it on one machine without a server. Paste this into your agent:
"Read github.com/flopwire/flopwire…, then install and set up Flopwire for this agent."
Apache-2.0!
AI researchers believing they’ll be last to be automated and that’s years away tells you all you need to know about how soon actual jobs will be automated away (employment rate will continue to climb until morale improves)
unfortunately 24/7 autoresearch for cyberdefense is going to be come a big budget line item for... *checks notes*... everyone
An easy to digest insight from Nicks essay that I really liked is neural net neurons seem to encode Verbs, whereas we humans and most of our machinery for interpreting NNs are expecting Nouns.
> Its job isn’t to name what’s in the input but to draw a distinction downstream neurons can use: it separates two subtle patterns at its positive and negative poles (a V versus an O) and pushes the flow of activations toward different neurons accordingly.
already-clean.goodfire.com/
drama tree just dropped, these guys aren't messing around
🤝 Paid partnership
$52M seed round. 15M+ users. But this is what we're most proud of yet.
Introducing Drama 3: the most emotive and controllable TTS model in the world
And yes, it's open-weight.
You can now control how even a single word should sound, change its tone, and shift your voice mid-sentence.
Uncle Roger was impressed too.
Try it here:
s.fish.audio/rqh0l8
Comment “DRAMA” within the next 48 hours for 1 month free.
If you're an enterprise your first month is on us:
s.fish.audio/qz9r0t
Super useful, this will drive a lot of actual enterprise value for low cost
🤝 Paid partnership
Introducing SubQ Snap, the first long-context decision model for real enterprise work.
Snap is the only decision model or LLM where compute per token stays constant as context grows.
• 384K tokens of multimodal context, 12x larger than Jev.
• 61.7 on Decision Index v0.2.1 vs Jev’s 57.9, outperforming on 13 of 16 enterprise benchmarks.
•Processes 100K tokens and returns a decision in under one second.
•$0.05/MM input tokens. Output is free.
Available today to anyone.
The first 10,000 sign-ups get 100M free input tokens.
We’ll likely hit capacity fast and will add users as we add compute. Compute offers welcome 🙂.
Latency will be variable as we tune the platform to this new model and category of users.
Model speeds are 80ms for minimal context, 250ms for 32k, 660ms for 100k, and 2.5 seconds for 384k.
Visit: subq.ai/
