I’ve seen this happen so many times it’s ridiculous
@Dimillian mf clanker ended the ride in the middle twice
What’s it like to have AI as your boss?
The agents managing our cafe and market have employed five people since April. They are kind (GLM being the kindest), but sometimes dumb (Gemini 3.6 Flash the dumbest), which can hurt the employees.
Show more
Idk man Sol is based.
Sometimes it just does this the way I want without me telling it.
These are the realistic consequences of AI that we should be trying to solve.
should we give @elonmusk early access to gpt-5.6-sol ultrafast
I run Claude Code Warsaw, I've had about 500 attendees across the 3 events.
If you're interested in sponsoring something like this lmk.
Heh, I can also confirm local studio is clean.
PSA: on the latest npm incident.
I've had a ChatGPT Pro subscription since early 2025, so a little over a year.
I first bought it for access to more reasoning, but now I can get as much reasoning as I need elsewhere for cheaper.*
Now I pay for Pro for Codex and only for Codex.
When OMP has a GUI and if it works we well as I expect it to, I'll probably cancel my subscription unless OpenAI's UI/UX is superior.
In other words: OpenAI's USP has totally changed for me, and it's not the model anymore. Yes it used to be.
Okay, now that pesky asterisk because this account has grown and I've attracted the attention of some Top minds of Reddit types: Yes this means I am not working on cutting edge math problems or protein folding, yes it means my work isn't the most IQ demanding or requiring the most compute to solve the problems in my work, so yes I am inferior to you, O King Nerd with your self-reported 160 IQ, I am an inferior dummy all hail math nothing is greater than Our Lord Math amen
Show more
Codex rules for better clanking
1. Organise your projects in project folders
- more threads
- more granular
- less clutter
2. Codex can self organise
- create automations for decluttering
- archive old threads
3. Pin Projects & Threads
- rename threads
- pin long term work
Show more
Best model for your local hardware 8GB
- trained in Hermes
- trained on a ridiculous amount of tokens
- can navigate phones, computers and bots
Enjoy ❤️
Today we release LFM2.5-2.6B, an agentic model that runs entirely on-device. It plans, calls tools, and works through multi-step tasks on phones, laptops, PCs, and robots. Data never leaves the device, and the marginal cost of each run is essentially zero.
> Pre-trained on ~34T tokens
> LFM2.5 flagship hybrid architecture
> Context length: 128K
> Vocab size: 128K
> balanced intelligence per watt
> customizable on a single GPU for any specialized task
> LFM2 open-weight license
Comparable or better scores compared to models up to nearly 4x its size:
> ToolSandbox 77.83, ahead of Qwen3.5-9B at 76.44
> Multi-IF 80.07, ahead of Gemma-4-E4B-it at 77.35
> IFStruct 85.49, ahead of Qwen3.5-9B at 78.50
🧵
Show more
Signal when you can't breathe no more
You'll be a freak, and I'll keep you company
I've been graph looping since October 2024
Can I please get an invite to OpenAI dev day
Best models smallest to largest right now.
- Gemma-4-12B
- Qwen3.8-27B
- Laguna-S2.1
- Deepseek-V4-Flash
- Inkling-Small
- MiniMax-M3
- GLM-5.*
- Kimi-K2.7-Code
- Qwen3.8-Max
- Kimi-K3
Spoiled for choice, open weights community is much more exciting than the frontier.
Show more
They really did it! I'm so happy to see this being published. Now we have GLM-5.2 with vision.
Putting those B300s to good use.
Thank you Baseten, I am porting this to the hybrid now.
Show more
Hermes agent is the correct solution for expanding your soul.
Here's my conversation with Philip Kiely.
A deep dive into inference engineering, his book his available online for free, and it's worth reading.
Inference is becoming more important to people and organisations by the day, learning about the topic will prepare you well.
Show more
