nlpnerd
Re:Astra can read and write weights to improve Qwen
on Tuesday November 03, 2026 @01:24AM (#50348736)
Yeah, what's with researchers making attention grabbing claims like startup founders these days?
DoorDash said it's only received 'several hundred' orders from AI agents
Before you fine-tune, compile: lessons from a payment-risk experiment
DeepSeek Unveils Open-Source Tool to Help Cut AI Operating Costs
We fired the frontier model from two AI pipelines. Here are the receipts
Scaledown.ai Launches Constrained Decoding Models
Re:Launch HN: Agent37 (YC F26) – Low-cost agent sandboxes for running ...
on Monday October 12, 2026 @02:23PM (#50060620)
What about Maritime.sh (which is also in YC)? I swear everybody in SF is building the same stuff. Lol
Re:Introducing SubQ Snap: A Long-Context Decision Model
on Monday October 12, 2026 @02:22PM (#50060595)
Doesn't OpenAI's decision API support up to 1M context on Luna?
Re:Typesafe AI raises $870M at $7.5B
on Saturday October 10, 2026 @01:04AM (#50029738)
Because training your own model requires substantial effort around data annotations, and then dealing with data drift etc in production.
Re:Typesafe AI raises $870M at $7.5B
on Saturday October 10, 2026 @12:56AM (#50029700)
One may consider it "rigidity" but many application backends (including LLM powered ones) never needed the full unconstrained capabilities of an LLM. What people needed was the zero-shot capabilities over tasks as opposed to have to train a classifier, span model, etc for every task.
Constraints can be an advantage in many systems. Consider typed and untyped programming languages. Many advantages with typed languages over the latter in terms of development and efficiency.
Re:Typesafe AI raises $870M at $7.5B
on Friday October 09, 2026 @03:43PM (#50025702)
Yeah, agreed. A model or primitive on its own has no moat and frankly limited value. The paradigm behind "System One" models on the other hand is potentially huge.
https://seldon-ai.com/blog/fronter-llms-are-semantic-interpr...
Re:Typesafe AI raises $870M at $7.5B
on Friday October 09, 2026 @03:00PM (#50025175)
This is the difference between a "hot" company in a good ecosystem like SF. Yeah, the funds can take 2-3 months to do their due diligence. By the time it's done, the round has closed, and then what good is the due diligence?
Re:Typesafe AI raises $870M at $7.5B
on Friday October 09, 2026 @02:58PM (#50025155)
You are assuming that the VCs have done their due diligence. For a "hot" company like Typesafe AI, most likely little due diligence was done. That's the way it's played.
Semantic Compute: From Interpreters to Compilers
The Shift from Models to Compound AI Systems
Skill-Guided Mining and Compilation of LLM Agent Traces
AI applications need better primitives than text generation
Re:Jev in 25 Lines of Python
on Wednesday September 23, 2026 @03:37PM (#49821383)
This is basically Temu Jev
Re:Snorkel AI triples valuation to $3.5B as demand for AI training dat...
on Tuesday September 22, 2026 @11:43PM (#49811394)
Yet another "AI data lab" that is really just running data labelling operations for the frontier labs. I do wonder where this business is going in the long term.
Snorkel AI triples valuation to $3.5B as demand for AI training data booms
Jev's Architecture Unmasked
The Great Unbundling of the LLM
Re:Understanding Inference and the Economics of Enterprise AI
on Monday September 07, 2026 @03:49AM (#49595240)
How real is this? That people are taking inference costs seriously enough to invent a new line item "AI COGS"?
Understanding Inference and the Economics of Enterprise AI
The Shape and Feel of the Post-AI Data Stack
Running a Software Factory Efficiently at Uber Scale
Observational reconstruction, then the inverse problem
Re:Stripe Clinches over $7B Deal to Buy AI Firm OpenRouter
on Monday August 17, 2026 @05:27AM (#49328334)
Except there are like a dozen Openrouter clones out there with nowhere near the same traction. Not as easy as people believe to build a successful startup. The team probably had the right experience and advantage from building OpenSEA.