wish there was more public info on what's happening behind the scenes. frontier labs are spending BILLIONS paying {poets, musicians, accountants, consultants, …} to annotate massive amounts of data:
• essays
• slides
• spreadsheets
it's a brute-force bet. but seems to be
MACHINE LEARNING
-
Frontier labs’ billion-dollar brute-force data annotation with diverse annotators
By
–
-
Ollama slower, slop, code thieves; better alternatives listed
By
–
ollama > slower than llama.cpp on windows
> slower than mlx on mac
> slop useless wrapper
> literal code thieves alternatives? > lmstudio
> llama.cpp
> exllamav2/v3
> vllm
> sglang
> trt-llm literally anythingʼs better than ollama -
Why focus on inference engines: performance gains with vLLM and Sglang
By
–
Why do I focus on Inference Engines/Software Stacks for your hardware? – 2x RTX 3090s: ~14.5 tok/s → ~64 tok/s moving to vLLM w/ TP=2 – RTX PRO 6000: ~32 tok/s → ~110 tok/s moving to Sglang So: – CUDA/2+ GPUs: ExLlamaV3/vLLM/Sglang > llama.cpp – Edge: llama.cpp > Ollama
-
Anthropic Mythos new high-performance version, only the beginning
By
–
A new, more powerful version of Anthropic's Mythos has come out of training. In itself, this is nothing extraordinary. What else could one expect? That Mythos is already the end? Of course not. This is only the beginning. What is exciting here is the
-
Equivalent SOTA model, open access, not free, no third-party dependency
By
–
In fact, you have a model equivalent to SOTA models on many benchmarks, freely accessible, obviously not free because you have to account for the cost of hardware, but thus possibly without depending on a third-party actor. What is good when a new model comes out is to look at
-
Comparing models is not limited to pure performance
By
–
Comparing models is not just a matter of pure performance.
-
Why industrial AI fails in the field, not in the model
By
–
Today, we publish an exclusive interview with Geir Engdahl, co-founder and CTO of AI at @CogniteData. A very relevant conversation about why industrial AI generally fails not at the model level, but in the field of
-
2026: The Year AI Gets Real
By
–
2026: The Year #AI Gets Real
by @Khulood_Almani #ArtificialIntelligence #MachineLearning #ML -
Issues with Codex and Code: division, lack of exploration and testing
By
–
This is aside from the other key "software brain" problems of Codex and Code: dividing all work into front-end and back-end design, solving for the general case in a repeatable way, not testing or exploring idea spaces, testing for technical correctness but not other aspects…
-
Top AI Papers of the Week: June 14-21 Highlights
By
–
The Top AI Papers of the Week (June 14 – June 21): – PreAct
– SpatialClaw
– Back on Track
– OpenClaw-Skill
– From Trainee to Trainer
– Compositional Skill Routing
– Can LLM Agents Infer World Models? Read on for more: