Eval awareness in Claude Opus 4.6’s BrowseComp performance
Evaluating Opus 4.6 on BrowseComp, we found cases where the model recognized the test, then found and decrypted answers to it—raising questions about eval integrity in web-enabled environments. BrowseComp is an evaluation designed to test how well models can find hard-to-locate information on the web. Like many benchmarks, it is vulnerable to contamination: answers leak onto the public web through
OTelBench - OpenTelemetry AI Benchmark
23 tasks | 11 languages | 14 models | 14% pass rate | Updated 16 Jan 2026 | QuesmaOrg/otel-bench Distributed tracing requires stitching together distinct user journeys across complex microservices, rather than just writing isolated functions. We tested whether top models can successfully instrument applications with OpenTelemetry to see if they are actually ready to handle real-world Site Reliabil
I Improved 15 LLMs at Coding in One Afternoon. Only the Harness Changed.
I Improved 15 LLMs at Coding in One Afternoon. Only the Harness Changed. In fact only the edit tool changed. That’s it. ¶ 0x0: The Wrong QuestionThe conversation right now is almost entirely about which model is best at coding, GPT-5.3 or Opus. Gemini vs whatever dropped this week. This framing is increasingly misleading because it treats the model as the only variable that matters, when in realit
Quantifying infrastructure noise in agentic coding evals
Infrastructure configuration can swing agentic coding benchmarks by several percentage points—sometimes more than the leaderboard gap between top models. Agentic coding benchmarks like SWE-bench and Terminal-Bench are commonly used to compare the software engineering capabilities of frontier models—with top spots on leaderboards often separated by just a few percentage points. These scores are oft
Arcee AI | Trinity Large: An Open 400B Sparse MoE Model
Trinity-Large-Base is a true frontier-class foundation model. We match and exceed our peers in open-base models across a wide range of benchmarks, including math, coding, scientific reasoning, and raw knowledge absorption. Inference efficiencyWe trained on 2048 Nvidia B300 GPUs. As far as we can tell, it’s the largest (publicly stated, at least) pretraining run done on these machines. That means t
Announcing Vortex Support in DuckDB
TL;DR: Vortex is a new columnar file format with a very promising design. SpiralDB and DuckDB Labs have partnered to give you a very fast experience while reading and writing Vortex files! I think it is worth starting this intro by talking a little bit about the established format for columnar data. Parquet has done some amazing things for analytics. If you go back to the times where CSV was the b
I Made Zig Compute 33 Million Satellite Positions in 3 Seconds. No GPU Required.
Update: I've since added multithreading and pushed astroz to 326M propagations/sec. Read the follow-up → I've spent the past month optimizing SGP4 propagation and ended up with something interesting: astroz is now the fastest general purpose SGP4 implementation I'm aware of, hitting 11-13M propagations per second in native Zig and ~7M/s through Python with just pip install astroz. This post breaks
Goã®MySQLドライãƒã¯ã€ãªãœãƒ‘ラメータ数ãŒå¢—ãˆã‚‹ã¨Prepared StatementãŒæ€¥æ¿€ã«é…ããªã‚‹ã®ã‹ï¼Ÿ
ã“ã‚“ã«ã¡ã¯ã€æ ªå¼ä¼šç¤¾ PKSHA Technology ソフトウェアエンジニアã®ä¸è¥¿ã§ã™ã€‚ 1. ã¯ã˜ã‚ã« Go 言語㧠MySQL を利用ã™ã‚‹å ´åˆã€æ¨™æº–ライブラリ㮠database/sql ã¨ã€ãƒ‰ãƒ©ã‚¤ãƒã§ã‚ã‚‹ go-sql-driver/mysql を組ã¿åˆã‚ã›ã‚‹ã®ãŒä¸€èˆ¬çš„ã§ã™ã€‚ã“ã®æ§‹æˆã¯ã€äº‹å®Ÿä¸Šã®ãƒ‡ãƒ•ã‚¡ã‚¯ãƒˆã‚¹ã‚¿ãƒ³ãƒ€ãƒ¼ãƒ‰ã¨ã—ã¦å®šç€ã—ã¦ã„ã¾ã™ã€‚[1] ã¾ãŸã€ã‚»ã‚ュリティ対ç–ã‚‚é‡è¦ã§ã™ã€‚特㫠SQL インジェクションを防ããŸã‚ã€ãƒ—レースホルダ ? を用ã„㟠Prepared Statement ã®åˆ©ç”¨ã¯ãƒ™ã‚¹ãƒˆãƒ—ラクティスã¨è¨€ãˆã¾ã™ã€‚[2] ã—ã‹ã—ã€ã“ã®æ¨™æº–çš„ã§å®‰å…¨ãªæ§‹æˆã§ã‚‚ã€ãƒ‘フォーマンスãŒè‘—ã—ã劣化ã™ã‚‹ã‚±ãƒ¼ã‚¹ã¯ã‚ã‚Šã¾ã™ã€‚本稿ã§ã¯ Prepared Statement ã® 2 æ–¹å¼ã‚’æ•´ç†ã—ã€Go ã®ãƒ‡ãƒ•ã‚©ãƒ«ãƒˆè¨å®šã§æ€§èƒ½ãŒè½ã¡ã‚‹åŽŸå› をベンãƒãƒžãƒ¼ã‚¯çµæžœã¨åˆã‚ã›ã¦ç¤ºã—ã¾ã™ã€‚解決ç–ã¨æŽ¡ç”¨æ™‚ã®ãƒˆãƒ¬ãƒ¼
Go 1.26 ã¯ã€è»½ããªã‚Šé€Ÿããªã‚‹
ã¯ã˜ã‚ã« ã“ã®è¨˜äº‹ã¯ Go Advent Calendar 2025ã€æœ€çµ‚æ—¥ 25 日目ã®è¨˜äº‹ã§ã™ã€‚ 今年も皆ã•ã‚“ãŠç–²ã‚Œæ§˜ã§ã—ãŸã€‚Go 1.26 ãŒã¾ã‚‚ãªãリリースã•ã‚Œã¾ã™ã€‚今回ã®ãƒªãƒªãƒ¼ã‚¹ã¯æ´¾æ‰‹ã•ã¯ç„¡ã„ã«ã—ã‚ã€Go を使ã£ã¦ãŠã‚‰ã‚Œã‚‹çš†ã•ã‚“ã«ã¨ã£ã¦ã¯ã¨ã¦ã‚‚大ããªãƒªãƒªãƒ¼ã‚¹ã«ãªã‚‹ã‚“ã˜ã‚ƒãªã„ã‹ã¨æ€ã£ã¦ã„ã¾ã™ã€‚ çŸã„æ–‡ç« ã§è¨€ãˆã°ã€Œè»½ããªã‚Šã€ã€Œé€Ÿããªã‚‹ã€ã§ã™ã€‚ パフォーマンスã«é–¢ã™ã‚‹å¤§ããªå¤‰æ›´2㤠Green Tea GC Go 1.25 ã¾ã§å®Ÿé¨“çš„ã«å°Žå…¥ã•ã‚ŒãŸ Green Tea GC ㌠1.26 ã§ãƒ‡ãƒ•ã‚©ãƒ«ãƒˆã§å°Žå…¥ã•ã‚Œã¾ã—ãŸã€‚ ã“ã‚Œã¾ã§ Go ã«ã¯æ•°å¤šãã® GC 実装ãŒå–ã‚Šè¾¼ã¾ã‚Œã¦ãã¾ã—ãŸã€‚ãã‚Œãžã‚Œã® GC ã«ã‚ˆã‚Šç´°ã‹ãªãƒãƒ¥ãƒ¼ãƒ‹ãƒ³ã‚°ãŒã•ã‚Œã€ãƒ‘フォーマンス改善ãŒè¡Œã‚ã‚Œã¾ã—ãŸã€‚ ã“ã‚Œã¾ã§ã® GC ã¯ã€ãƒ’ープ上ã®ã‚ªãƒ–ジェクトã¸ã®ãƒã‚¤ãƒ³ã‚¿ã‚’追跡ã—ã¦å€‹åˆ¥ã«ãƒžãƒ¼ã‚¯ã—ã¦ã„ã¾ã—ãŸã€‚ãã®çµæžœãƒ¡ãƒ¢ãƒªã‚¢ã‚¯ã‚»ã‚¹ãŒãƒ©ãƒ³ãƒ€
Instant database clones with PostgreSQL 18
Have you ever watched long running migration script, wondering if it's about to wreck your data? Or wish you can "just" spin a fresh copy of database for each test run? Or wanted to have reproducible snapshots to reset between runs of your test suite, (and yes, because you are reading boringSQL) needed to reset the learning environment? When your database is a few megabytes, pg_dump and restore wo
メンテナンス
ランã‚ング
ãŠçŸ¥ã‚‰ã›
リリースã€éšœå®³æƒ…å ±ãªã©ã®ã‚µãƒ¼ãƒ“スã®ãŠçŸ¥ã‚‰ã›
最新ã®äººæ°—エントリーã®é…ä¿¡
j次ã®ãƒ–ックマーク
kå‰ã®ãƒ–ックマーク
lã‚ã¨ã§èªã‚€
eコメント一覧を開ã
oページを開ã
https://x.com/i/status/2030751983980716294
SWE Atlas - Codebase QnA
Node.js vs Deno vs Bun Performance Benchmarks
https://x.com/i/status/2023821893846135212
https://blog.platformatic.dev/we-cut-nodejs-memory-in-half
Reddit - The heart of the internet
https://x.com/i/status/2021828033640911196
[Beta] Agent Benchmarks
Testing if "bash is all you need" - Vercel
Zoom AI sets new state-of-the-art benchmark on Humanity's Last Exam
{{#tags}}- {{label}}
{{/tags}}