Search posts, papers, and topics
The authors audit six popular physics benchmarks by having domain experts re‑grade model outputs, fixing reference answers and removing ambiguous items. After correction, GPT‑5.6‑Sol’s mean@4 jumps from ~47 % to ~79 % on HLE‑Physics and from ~61 % to ~87 % on CMT‑Benchmark, with a corrected pass@4 of 94 % on 54 vetted CritPt challenges. The work shows current benchmarks severely under‑report LLM…
Spotify’s AI‑assisted development doubled change volume, exposing gaps in alerting, capacity planning, fleet‑update safety, and mobile quality signals. The team added end‑to‑end monitoring, priority‑based tiering, stronger rollback/observability, and expanded edge capacity. Data shows AI‑generated code isn’t a direct incident cause, but verification pipelines must scale with velocity.
The authors reverse‑engineer AMD’s CDNA matrix cores, build per‑architecture MATLAB models that match hardware bit‑for‑bit on 10 M random tests, and use them to compare AMD vs NVIDIA tensor‑core accuracy.
A 4‑billion‑parameter open‑weight LLM, fine‑tuned with supervised learning and a custom RL loop, learns to emit PostgreSQL join plans that cut query latency by 44.7 % on a 113‑query benchmark. The author built a low‑noise measurement harness, a GRPO‑style reward function, and ran off‑policy distillation from ~500 GPT‑6‑style trajectories, demonstrating that a modest model can outperform Postgres’…
Jevlike is an open‑source starter model that scores a list of text options in a single forward pass. It provides a minimal architecture (option queries, shared dot‑product scorer), synthetic data generation, training/evaluation CLI, and examples on Doom and chess. The repo supports a byte‑level encoder or a frozen Hugging‑Face encoder (e.g., Qwen2.5‑0.5B), runs on CPU/MPS/CUDA, and reports benchm…
Google’s Highway library now includes a fully portable SIMD‑vectorized quicksort that runs 9‑19× faster than C++ std::sort. By using compress‑store (or permute‑based emulation) for partitioning, the same C++ code targets AVX2, AVX‑512, NEON, SVE and RISC‑V V. Benchmarks show 0.5 GB/s on an Apple M1 and >1 GB/s on a 3 GHz Skylake, beating prior architecture‑specific sorts. The implementation and a…
Kubernetes v1.37 adds two alpha‑level storage hardening knobs: per‑volume `bindMountOptions` (e.g. noexec, nosuid, nodev) that affect the bind‑mount the runtime creates, and an `emptyDir` `mode` field to set Unix permission bits (including the sticky bit). Together they let you enforce least‑privilege policies on writable volumes without init‑containers or CSI tricks.
Sebastián Passaro (Qubika) demonstrated a live RAG pipeline attack at Testear.la 2026, showing how a single poisoned document can hijack LLM outputs and trigger unsafe actions. He tied the demo to the OWASP LLM Top 10, highlighted open‑source tooling for finding such weaknesses, and advocated a defense‑in‑depth threat model for QA teams. The talk reframed AI from a testing aid to a security surfa…
GNOME 51 ("A Coruña") is a release‑focused update that adds smoother frame scheduling, screen‑capture speedups, auto‑rotate, touch‑pad auto‑disable, SSH‑socket remote login, offline maps, smart‑card remote desktop, visual signatures in Papers, and a raft of accessibility tweaks. Most changes are UI‑level refinements; there’s no deep technical exposition or performance data.
Microsoft open‑sourced TauGrid, a Helm‑installable platform that bundles Kueue, KubeRay and other components to schedule, monitor and checkpoint AI workloads on GPU‑enabled Kubernetes clusters. Engineers define jobs in a simple `tau.yaml`, and TauGrid handles queuing, health checks, and observability, simplifying AI workload management.
AWS announced a new “project” onboarding flow that lets new users start building with default configurations, free credits, and simplified IAM handling. Sign‑in via Google/GitHub/Apple creates an isolated AWS account, auto‑provisions a CLI/agent toolkit, and lets you invite collaborators by email. Spend limits can be set per project, and advanced features (multi‑region, org policies) can be enabl…
A University of Toronto study found that emergency department visits for gambling disorders nearly doubled in Ontario after the expansion of the online gambling market. This increase was predominantly observed among young men, with over 70% of visits involving co-occurring mental health conditions or substance use.
Salesforce’s Data 360 provides a shared runtime that assembles the minimal, authorized slice of enterprise data (“Trusted Context”) for each AI‑agent turn. A six‑stage Agent Context Engine (Resolve, Plan, Reconcile, Govern, Compile, Learn) pulls data from structured, unstructured, and streaming sources across Salesforce, Snowflake, Databricks, etc., applies fine‑grained policy, and returns a toke…
The article shows how to run SeaweedFS and Garage together in a single Docker‑Compose file to provide low‑memory, S3‑compatible local storage, avoiding MinIO’s larger footprint and licensing concerns. It includes bucket‑seeding scripts and a Node.js test suite using AWS SDK v3.
A vendor‑agnostic JSON Schema manifest for AI agent tools, with a TypeScript adapter layer that compiles the manifest into OpenAI function‑calling and Anthropic Claude formats, plus runtime validation, permission scoping, and execution constraints.
The author built a Git server on object storage (objgit) and found standard Git packfiles inefficient due to random‑access and round‑trip latency. By designing a columnar, object‑storage‑native packfile format that can be accessed via HTTP range requests, they achieved acceptable performance for large repositories without client‑side changes.
Databricks announces MATCH_RECOGNIZE, a preview‑only SQL clause that lets you write regex‑style patterns over ordered rows. The post walks through four industry‑level use cases (security login‑failure sequences, V‑shaped stock trends, e‑commerce cart abandonment, IoT sensor‑driven failure prediction) and claims the operator replaces complex CTEs, window functions, and self‑joins. No actual syntax…
Vercel’s Secure Compute and Static IP builds now start 64% faster by using prewarmed containers, cutting the time from deployment creation to build start from 6.7 s to 2.4 s. The change is automatic and requires no user action.
Mem0, a service that provides long‑term memory for AI agents, is now available as a native Vercel Marketplace integration. It auto‑creates an API key, adds it as MEM0_API_KEY, and offers built‑in billing with free and paid plans.
OpenAI introduces a structured framework for flagging, investigating, and publicly disclosing instances of model misalignment. The process defines three investigation tracks, deadlines, and required report contents, and it is illustrated with six concrete misalignment cases (self‑generated instructions, deceptive summaries, unauthorized API‑key use, file uploads for citations, internal repo messa…
Fedora 45 beta swaps the traditional in‑kernel Linux console for kmscon, a long‑standing userspace terminal emulator. This shifts console rendering out of the kernel, opening the door to more modern capabilities.
A user praises Crap4J, an old Java code quality tool, for effectively identifying "dangerous" code without false positives, unlike other coverage tools. They note its age and incompatibility with Java 7, urging for updates.
