Loading activity…
@dimitrispapail
Researcher @Microsoft | Prof @UWMadison (on leave) | babas of Inez Lily.
AI
DimitrisPapail calls Lazarus a training-free weight-editing backdoor; the available posts include no technical demonstration.
AI
His account describes a clash over scaling and research after the group’s move to Microsoft AI.
AI
The reported MATH-500 score rose from 7.9% to 48.1% with the same sampling settings.
AI
ChangJonathanC proposes layer looping; quoted poster @RhLm2337 reports a shared neuron index but calls the larger model “qwen3.8 27b.”
AI
@irenetrampoline reports their team’s award and an opportunity to develop Edge Medicine with JupyterHealth.
AI
A post covering GitHub Universe 2026 also quotes a projects-related feature as spinning up as many agents as needed.
AI
Their post frequency and “shitpost rate” will likely decrease, scaling01 says; their views remain their own.
AI
The announcement says Qwenputer runs C and MicroPython during forward passes, without an external runtime.
XAI
OpenAI says its claimed refutation of the Exponential Time Hypothesis was formalized in Lean and made public.
AI
The zero-loss weight changes are Litman's reported result; moyix questions his supply-chain zero-day framing.
AI
In a post, kushal1t claims the framework boosts GPU utilization by packing independent workloads into one kernel.
AI
Its developer says Flint exceeds 125 tokens per second on an M4 Pro MacBook Pro at short contexts.
XAI
One post argues an agent could run programs and make slides without the desktop layers built for human users.
AI
Anthropic reports 17 of 211 lab-tested designs met binding criteria, but none were tested for toxin neutralization.
AI
A post says the updates come from the team behind one of its author's open-source support initiatives.
AI
Meta says it runs at practical speeds on a Mac or PC with one consumer GPU at 4-bit precision.
AI
A commenter says the issue boils down to a lack of empathy with the recipient's state of mind.
AI
A user suggests 10× the input buys roughly a constant increment of output in both relationships.
AI
A user jokes that one frontier lab is preparing a better model while others are reportedly releasing better successors.
AI
A user recommends linked Anthropic research as a grounded, quantitative and qualitative case study of that change.
