Daily Tech News - 2026-09-17

AI’s next test is operational: making powerful systems cheaper, faster, and reliable enough to trust.

Across September 17’s coverage, infrastructure took center stage. AMD published guidance for reproducing its MLPerf Inference v6.1 submissions on Instinct GPUs, while CoreWeave reported deploying a multi-rack NVIDIA Vera Rubin NVL72 cluster for agentic workloads. Raptor accelerators’ planned integration with NVIDIA’s NVLink fabric added another approach to scaling inference. Alongside these hardware developments, analysis of DeepSeek-v4.1 Flash’s KV cache compression highlighted the push to reduce inference’s memory demands.

For developers, the focus shifted toward turning AI experiments into working systems. A JetBrains developer diary explored retrieval-augmented semantic code search, while OpenClaw deployment guidance addressed taking agents into production. One developer’s account of having Gemini train a replacement for $9 offered a small-scale counterpoint to the infrastructure race, raising questions about how cheaply specialized tasks can be handled.

That acceleration puts greater weight on verification. Coverage of vLLM performance and accuracy checks, provably correct Rust with Verus, and test suites as a foundation for reviewing agent-generated code all pointed toward the same challenge: establishing confidence as software changes faster. A report on four SSRF bypasses underscored the limitations of security blocklists.

Meanwhile, platform engineering continued tackling everyday friction. Red Hat explored a unified CI/CD portal and faster performance-test setup; Debezium introduced a server rewrite supporting native builds with Quarkus and Mandrel. ClickHouse highlighted query and text-search improvements, and Kubex open-sourced tooling for observing processes on shared GPUs.

Modular’s opening of Mojo compiler contributions rounded out a day in which ambitious computing advances met the practical work of making them usable.

Featured Articles

Reproducing AMD MLPerf Inference v6.1 Submission Results

Reproducing AMD MLPerf Inference v6.1 Submission Results# This blog shows you how to reproduce AMD submission results for MLPerf Inference v6.1 on AMD Instinct MI355X, MI350X, and MI350P GPUs using se...

Keywords: amd mlperf, benchmarks mlperf, mlperf mi355x_deepseek_r1_inference_6, amd specifically, benchmark suite, benchmarks run, performance benchmark, amd submission, running benchmark, reproducing amd
Source: rocm.blogs.amd.com

Supersonic Subatomic Debezium Server: Native builds with Quarkus and Mandrel

The Debezium 3.7.0.Beta2 release comes with a lot of changes, one of which is a complete rewrite of Debezium Server to support native builds with Quarkus and Mandrel. In this article, we’ll walk throu...

Keywords: debezium platform, build debezium, jvm runtime, debezium libraries, debezium native, builds java, java platform, run debezium, debezium server, java applications
Source: debezium.io

Build a unified CI/CD control plane with Red Hat Developer Hub

If your engineering teams waste time context-switching between Red Hat OpenShift, Argo CD, Tekton, and Quay, your platform needs a unified developer portal. Building on the previous article in this se...

Keywords: openshift pipeline, openshift pipelines, openshift argo, pods deployments, openshift console, deployments pipeline, managed openshift, switching openshift, pipeline argo, status openshift
Source: developers.redhat.com

Building a RAG Pipeline for Semantic Code Search: A Developer Diary and Field Notes

JetBrains AI Supercharge your tools with AI-powered features inside many JetBrains products Building a RAG Pipeline for Semantic Code Search: A Developer Diary and Field Notes Part 1: Parsing, chunkin...

Keywords: coding agents, agent complete, agent needs, agent driven, tools ai, reasoning agent, ideas agentic, agentic code, snippets agent, jetbrains ai
Source: blog.jetbrains.com

How to Deploy OpenClaw Agents to Production: Best Practices

The rapid advancement of artificial intelligence and open-source frameworks has positioned AI agents as crucial components in modern software architectures. OpenClaw, a newer open-source AI agent fram...

Keywords: openclaw ai, openclaw agents, openclaw agent, deploying ai, ai deployment, architectures openclaw, deploying agents, openclaw production, agents deploying, agents automate
Source: collabnix.com

GitLab 19.4 release notes

GitLab 19.4 release notes On September 17, 2026, GitLab 19.4 was released with the following features. We are excited to recognize Jimmy, a Level 4 contributor, as this month’s Notable Contributor! Ji...

Keywords: supported gitlab, governance gitlab, managed gitlab, enable gitlab, administrators gitlab, dedicated gitlab, govern gitlab, internal gitlab, gitlab instance, gitlab dedicated
Source: docs.gitlab.com

Vinix – A modern operating system written in V

Apple Silicon Macs QEMU uses macOS hardware virtualization for the ARM64 desktop. brew install qemu git clone https://github.com/vlang/vinix cd vinix ./run-desktop-aarch64.sh 100 MB RAM usage on boot...

Keywords: macs qemu, qemu uses, vinix linux, install qemu, vinix kernel, qemu, vinix desktop, arm64 vinix, run vinix, kernel vinix
Source: vinix-os.org

DeepSeek-v4.1 Flash: Pushing the Limits of KV Cache Compression

DeepSeek-V4.1 Flash: Pushing the Limits of KV Cache Compression TL;DR When DeepSeek-V4.1 Flash was released, I thought it might just be a post-training iteration version... but after using it for a wh...

Keywords: performance deepseek, kvcache deepseek, deepseek v4, cache deepseek, deepseek v5, kv cache, deepseek optimizations, deepseek v3, compression deepseek, compress_kv_cache
Source: zartbot.github.io

I had Gemini train its own replacement for $9

I had Gemini train its own replacement for $9 I like to cook, and somewhere along the way that turned into an obsession with high-end chef's knives. So I scrape the Reddit threads where people argue a...

Keywords: comments products, product names, marking brand, benchmade label, brand text, product labeled, picking product, labels gemini, gemini label, steel names
Source: petervijeh.com

Developing provably correct Rust code with Verus

Many open-source and industry software projects, including several here at Amazon, are embracing the Rust programming language, since it provides performance and flexibility similar to that of the C p...

Keywords: rust implementation, rust programming, rust compilers, verus rust, guarantees rust, rust program, code rust, rust code, unsafe rust, using rust
Source: amazon.science

Hello and welcome to another ClickHouse newsletter! We’ve got another feature-packed ClickHouse release, with custom HTTP handlers, pipelined SQL, and Japanese/Chinese tokenizers for the text-index. E...

Keywords: clickhouse replication, replication clickhouse, postgres clickhouse, clickhouse postgresql, clickhouse cloud, roadmap clickhouse, availability clickhouse, postgresql clickhouse, queue clickhouse, replicates postgres
Source: clickhouse.com

The Golden Spike, and Resurrecting the Vale(n) Programming Language

There was an incredible moment on May 10th, 1869, when railroad builders finally reached their goal of joining the east coast rail network with the west coast rail network. In that moment, the first U...

Keywords: transcontinental railroad, rust languages, transcontinental compiler, languages rust, compiler rust, rust compiler, railroad, golden spike, rustc seamlessly, 1869 railroad
Source: verdagon.dev

Catching poor LLM performance and accuracy before deployment

vLLM is the leading open source inference and serving engine for LLMs. Averaging 64 commits a day with bi-weekly releases, the open source project goes through a significant amount of code changes rap...

Keywords: midstream benchmarks, vllm releases, regressions midstream, vllm release, llm vllm, upstream vllm, llms performance, downstream midstream, midstream releases, vllm llm
Source: developers.redhat.com

Frontier class ultra-low-latency rack-scale inference with Raptor XPUs and NVLINK Fusion

We announced that we are partnering with Nvidia to support rack-scale deployments of our 3D-DRAM enabled Raptor accelerators using the NVLink scale-up fabric on identical infrastructure to Nvidia’s Ve...

Keywords: nvidia rack, raptor nvidia, rack nvlink, gpu raptor, rack gpu, raptor gpu, raptor architecture, latency stacked, raptor workload, gpu racks
Source: d-matrix.ai

Intel Compute Stack Enables LEO For Nova Lake, Introduces Intel Portable ISA "PISA"

Intel Compute Stack Enables LEO For Nova Lake, Introduces Intel Portable ISA "PISA" Released today was the Intel Compute Runtime 26.35.39758.10 update for Windows and Linux systems. With this updated...

Keywords: opencl support, opencl implementation, intel compute, opencl leo, experimental opencl, apis opencl, opencl driver, opencl direct, opencl level, opencl
Source: phoronix.com

CoreWeave Deploys Multi-Rack Vera Rubin NVL72 Cluster: Hundreds of Rubin GPUs, 1.6 Tb/s Per GPU, and a No-Fee Archive Tier

CoreWeave has deployed a multi-rack NVIDIA Vera Rubin NVL72 cluster on CoreWeave Cloud, connecting hundreds of Rubin GPUs into a single scale-out environment aimed at agentic AI workloads. Each Dell-b...

Keywords: nvl72 architecture, nvl72 cluster, cpus nvlink, vera cpus, rack nvidia, nvidia vera, nvme coreweave, gpus coreweave, nvlink gpu, coreweave cloud
Source: storagereview.com

From hours to minutes: Optimizing Red Hat Developer Hub performance testing with immutable LDAP images

At Red Hat, our CI/CD pipelines are the heartbeat of our development process. However, as we scaled the Red Hat Developer Hub performance testing framework, we hit a wall. Our environment setup time w...

Keywords: performance testing, testing faster, performance framework, performance bottleneck, bottleneck runtime, database performance, faster troubleshooting, api leveraging, massive catalog, testing framework
Source: developers.redhat.com

Do engineers still need to understand how LLMs work?

Using AI agents for coding has become standard today. But not many people can say that the model can write that code. My guest today is someone who spent the last few years writing books on that topic...

Keywords: ai agents, ai agent, models agents, using agents, agents coding, ai writing, learning ai, using ai, ai api, ai coding
Source: newsletter.techworld-with-milan.com

Four SSRF Bypasses in Four Months

-5b301f10ddcd---4 crawled_date: 2026-09-17T08:51:07.035953+00:00 feed_url: https://itnext.io/feed published: Thu, 17 Sep 2026 08:56:32 GMT

Four SSRF Bypasses in Four Months Why Your Blocklist Was...

Keywords: ssrf bypasses, patch ssrf, ssrf vulnerability, proxy ssrf, ssrf protection, recommends curlopt_resolve, ssrf prevention, ssrf aware, ssrf, hop ssrf
Source: itnext.io

6 Ways Traditional API Design Has Changed Forever

6 Ways Traditional API Design Has Changed Forever J Simpson September 17, 2026 For over two decades, API design and architecture remained remarkably consistent. Perhaps it’s the outsized influence of...

Keywords: traditional apis, api architects, apis designed, traditional api, apis introduced, api designers, api agent, api architecture, apis adopted, apis used
Source: nordicapis.com

A decade of Gazette: a streaming broker that doesn’t store data

There’s something even more scary than live traffic in a streaming system: backfills. A new application almost never wants just the events arriving from that moment forward, it wants the history too,...

Keywords: streaming brokers, streaming broker, stream brokers, gazette streaming, streaming infrastructure, streaming systems, broker processes, broker synchronizes, synchronous broker, replication streams
Source: estuary.dev

Open Sourcing Kubex’s GPU Process Exporter: Gain Visibility in Your Shared GPUs

Open Sourcing Kubex’s GPU Process Exporter: Gain Visibility in Your Shared GPUs GPU sharing with NVIDIA hardware is becoming easier to adopt in Kubernetes but it hasn’t been easier to observe. Time-sl...

Keywords: kubex gpu, gpus kubernetes, gpu processes, gpu sharing, shared gpu, gpu containers, gpu process, share gpu, sharing gpu, shared gpus
Source: kubex.ai

Modular: Modular 26.6: Open compiler contributions, audio generation, and expanded model support

Modular 26.6 is our first release since open sourcing the Mojo compiler at ModCon. With Mojo 1.1, we’re taking the next step in that journey by opening the compiler to external contributions and movin...

Keywords: mojo compiler, modcon mojo, mojo builds, upgrade mojo, mojo install, compiler modcon, mojo changelog, 26 mojo, contributions modcon, building mojo
Source: modular.com

The Test Suite Is the New Code Review

Some mornings I watch a pull request move through our whole pipeline before I’ve finished my coffee. An agent opens it. A second agent reviews it. By the time I sit down, the only thing still running...

Keywords: pull requests, pipeline merge, merge slowest, merges pull, pipeline works, requests pull, agent pull, tests merge, letting pipeline, pipelines
Source: allen.hutchison.org

What Happens When You Ask ChatGPT a Question

What Happens When You Ask ChatGPT a Question You type a question into ChatGPT and hit send. A JSON body leaves your laptop as an ordinary HTTPS POST. The same POST, more or less, whether the box you t...

Keywords: chatgpt question, chatgpt hit, question chatgpt, chatgpt, ask chatgpt, typed chatgpt, chat completions, token sends, chat model, sends token
Source: manthanguptaa.in

Bonsai 2 27B: Near-Lossless Compression in a 9x Smaller Footprint

Introducing Bonsai 2 27B: Near-Lossless Compression in a 9x Smaller Footprint Two months ago, we released our first Bonsai 27B models and showed that a 27B-class multimodal model could be compressed e...

Keywords: ternary bonsai, alternatives bonsai, bonsai 27b, bonsai models, use bonsai, devices bonsai, introducing bonsai, lossless compression, release bonsai, capability bonsai
Source: prismml.com

hister

Your own search engine Hister is a private search engine for the pages you visit and the files you keep. It indexes their full contents so you can find information again from the web interface, termin...

Keywords: hister search, search browser, search provider, search configuration, search results, search indexes, search index, indexes search, indexing search, search engine
Source: github.com

upstrace 0.2.0

Upstrace When a data quality check fails, trace it upstream to the change that caused it. Upstrace profiles every column of a dbt project on every run, compares each day against its own history, walks...

Keywords: fault upstrace, upstrace dbt, upstrace fault, runtime upstrace, runs upstrace, behaves upstrace, upstrace runs, causes upstrace, analysis upstrace, upstrace runtime
Source: pypi.org

AI Reliability Engineering (AIRE): Dependable Humans | Solo.io | Solo.io

It’s a little after 5pm, and I’m about to wrap up for the day. As I’m starting to shut things down, I get a message from my boss: Jeff and Lisa are looking into some issues, the mortgage team is compl...

Keywords: warning signs, issue alerting, scenario alerts, alert fatigue, engineers alert, early warning, work risk, enhance reliability, alerting, carpool tonight
Source: solo.io

Bend

curl -fsSL https://bend-lang.com/install.sh | sh When using Bend:

run bend guide to learn it
use LAWS.bend to keep important rules
run bend PROOF.bend before committing
parallelize the c...
Keywords: bend compiles, bend ai, codebase bend, proof bend, bend proof, language bend, run bend, bend declare, fast bend, bend run
Source: bend-lang.com

How to find and debug agent failures your evals are missing

TL;DR: Predefined evals and monitors measure risks you already know how to specify. Production agents also develop failures nobody thought to encode: a tool receives the wrong argument, a retry loop a...

Keywords: evaluators maintain, evaluators experiments, evaluator catches, evaluation artifacts, failure patterns, evaluator checks, evaluators test, production evaluation, evaluator monitor, eval monitor
Source: arize.com

Daily Reading List – September 16, 2026 (#868)

Talked to a company today who asked me about some advice they got from another vendor. “Should we really use frontier models for everything at our company, or is that vendor delusional?” Smart compani...

Keywords: documentation ai, onboarding agent, kubernetes engine, google kubernetes, trusted documentation, ai maintenance, gke agent, trust agent, agents work, agents just
Source: seroter.com

How to build a Language Server Protocol (LSP) plugin for Claude Code

How to build a Language Server Protocol (LSP) plugin for Claude Code Content Marketing Manager Language servers give editors structured, real-time feedback: diagnostics, hover docs, autocomplete, and...

Keywords: language server, languages tools, language servers, build language, lsps language, existing language, code language, reuse language, server compiler, wrap language
Source: circleci.com

Postgres week in the Netherlands: PGDay Lowlands & Percona Live 2026

Last week, I spent three days in the Netherlands and gave two talks at two conferences: a lightning talk at PGDay Lowlands in Utrecht on Thursday, September 10, and a session at Percona Live in Amster...

Keywords: utrecht pgday, utrecht thursday, postgresql conference, amsterdam friday, amsterdam utrecht, utrecht went, days netherlands, utrecht, conferences lightning, dutch postgresql
Source: clickhouse.com

Harness Engineering: Why a Minimal CLAUDE.md and a Good Architecture Document Belong Together

The term Harness Engineering has established itself in recent months. The formula behind it is simple: Agent = Model + Harness. The harness is everything that makes up an AI agent, except the model it...

Keywords: harness engineering, process harness, harness operationalize, business harness, harness layers, ai unifiedprocess, harness layer, harness builds, model harness, build harness
Source: martinelli.ch

How and Where the Information You Give AI Travels

If you ask a coding agent to fix an authentication bug, very little appears to happen from where you sit. You write a sentence, wait a few seconds and get an answer back. The request the model receive...

Keywords: agents support, support agent, assistant agent, support assistant, agent queries, agent access, agent read, agent fix, coding agents, coding agent
Source: helmcode.com

The future of orchestration: Pine59’s journey to Airflow 3 on Google Cloud

The future of orchestration: Pine59’s journey to Airflow 3 on Google Cloud Piotr Wieczorek Lead Senior Product Manager, Google Jan Hajný Senior Data Engineer, Pine59 Operating large data pipelines req...

Keywords: airflow google, airflow cloud, traffic pipeline, data pipelines, google cloud, managed airflow, cloud, traffic computes, cloud heavy, airflow company
Source: cloud.google.com

Good Morning, Your Toaster Is Compromised

The following is a transcript of the 7am hour of Rise & Grind, a nationally syndicated breakfast news programme, first broadcast Thursday 17 September 2026. Transcription by OpenClaw-4.2. Portions of...

Keywords: fox breakfast, morning chip, chip toast, syndicated breakfast, fox chip, toasters morning, cartoon fox, toasters chip, grind chip, breakfast news
Source: nesbitt.io

What is AIOps?

AIOps combines AI and machine learning with observability to automate IT operations. Explore its core components, benefits, use cases, and challenges. Artificial Intelligence for IT Operations (AIOps)...

Keywords: infrastructure aiops, systems aiops, aiops analyzes, evaluating aiops, aiops analyze, services aiops, integrating aiops, aiops platform, aiops organizations, implement aiops
Source: databricks.com

Flet 1.0 – Build cross-platform apps in Python

Good-looking Python GUIs 150+ controls and services. Layouts, navigation, forms, and dialogs, with customizable colors, typography, and themes. Explore the controlsBuild beautiful web, desktop, and mo...

Keywords: controls python, python guis, button python, web flet, counter flet, stylesimport flet, python browser, page floating_action_button, flet build, flet studio
Source: flet.dev

Closing the loop: agentic evaluation for image editing foundation models

Why evaluating image editing models is both critical and challenging Instruction-based image editing is becoming a core capability of multimodal foundation models. Users can increasingly edit images s...

Keywords: editing evaluation, automate evaluation, editing models, ai tools, image editing, evaluation image, automating evaluation, improve models, edits evaluation, improve ai
Source: lambda.ai

Implementing a High-Performance Custom Diffusion Attention Kernel with FlyDSL

Implementing a High-Performance Custom Diffusion Attention Kernel with FlyDSL# Readers may be familiar with traditional Transformer models and their attention mechanisms. The traditional autoregressiv...

Keywords: flydsl kernels, kernel implementations, optimizing kernels, performance kernel, performance flexible, kv cache, kernel flydsl, kv caches, flydsl kernel, kernel highly
Source: rocm.blogs.amd.com

Show HN: AutoBot – live voice control for long-running AI work

AutoBot: a self-improving agentic harness that makes frontier AI better at finishing complex knowledge work. AutoBot achieved 18.5% higher task completion than the published OpenAI Sol Max baseline, s...

Keywords: work autobot, autobot, autobot achieved, execution autobot, autobot using, agent autobot, production autobot, autobot plus, autobot self, autobot combination
Source: github.com

Tilia—a new formatter for Haskell

Tilia—a new formatter for Haskell Published on September 16, 2026 Haskell remains a difficult programming language to format. The core parsing/printing machinery can be built relatively easily now tha...

Keywords: haskell formatters, formatter haskell, haskell parser, formatters operators, formatter fixities, haskell operator, ghc parser, specific haskell, fixities operators, haskell
Source: markkarpov.com

OpenTelemetry everywhere: Migrating a metrics platform at scale

Why we did this at all For most of the last decade our metrics pipeline ran on gostatsd, the open-source StatsD implementation we maintain. It primarily did two jobs: as sidecar on every host and the...

Keywords: infrastructure gostatsd, gostatsd aggregators, opentelemetry usage, pipeline opentelemetry, exploring opentelemetry, opentelemetry ecosystem, metrics services, statsd clients, gostatsd udp, statsd udp
Source: cncf.io

Pointers for your search career in 2026

My colleague Brian Pedersen recently released a report on current hiring trends. It’s a fascinating snapshot on what the market wants from new hires. How does this apply to you, the search professiona...

Keywords: search career, search professional, hiring trends, search relevance, applications search, participating search, search models, agentic search, diverse search, search diverse
Source: softwaredoug.com

Your GitHub Actions cache is the real bottleneck - Bitrise Blog

Namespace, Avrea, and Blacksmith are among the fastest GitHub Actions runner providers on the market right now. According to Avrea's August 2026 benchmarks, they all score within about 1% of each othe...

Keywords: slower github, slower builds, fastest github, fast builds, xcode benchmark, mac benchmark, cpu benchmark, github cache, warm builds, xcode workloads
Source: bitrise.io

How Uber Protects Against Retry Storms

How Uber Protects Against Retry Storms Senior Staff Engineer Principal Engineer Sr Software Engineer Introduction Retry storms historically impact business operations and brand trust. While retry conf...

Keywords: failures retry, downstream retry, handling retry, introducing retry, retry behavior, retry disturbance, retry storms, aggressively retrying, retry storm, allowing retry
Source: uber.com

Don't Make Job Referrals Public

Don't Make Job Referrals Public I am currently in the job market, so I receive some recruiter emails. Recently, I started seeing a slightly different kind of message. I posted in Hacker News' "Who wan...

Keywords: hired referral, job referrals, referrals work, referrals public, referral pays, successful referral, does referral, public referral, different referral, advertising referral
Source: blog.melashri.net

How we scope a Flutter MVP so founders get a first-value moment in weeks

Fuzzy MVP scope is how founders quietly lose weeks. Not in a dramatic rewrite – in the smaller places: debating screens that do not change the learning goal, polishing settings nobody will open in wee...

Keywords: flutter mvps, purpose flutter, flutter mvp, mvp scope, capabilities flutter, choose flutter, flutter specific, flutter, 2026 flutter, flutter android
Source: technotalkative.com

Pangram – AI detector for text and images

Detect AI-generated content with remarkable accuracy. Trusted by universities, schools, and enterprises worldwide. Proven the most reliable and accurate AI detector on the market by third party resear...

Keywords: ai content, ai writing, document ai, detect ai, detects ai, identifying ai, ai detection, distinguish ai, ai researchers, ai generated
Source: pangram.com

The Real Reason Coding Agents Get More Done

The Real Reason Coding Agents Get More Done Coding agents such as Claude Code and Cursor can do more than agents available inside most SaaS products. You can give a coding agent an issue, and after so...

Keywords: coding agents, saas ai, agent developers, agents coding, agents ai, coding agent, agents use, agents operate, api agent, agent applications
Source: builder.io

VQ-bench: a Composable Vector Quantization Framework

Before a vector database can search vectors, it has to store them. But storing high-dimensional vectors at full precision is quite expensive. Vector quantization (VQ) reduces the number of bits needed...

Keywords: existing quantizers, vectors quantize, vectors quantizer, vector quantization, quantization applications, implemented quantizer, building quantizer, quantizer implement, quantizers datasets, quantizers rarely
Source: pinecone.io

50 System Design Patterns Every Developer Should Know

Preparing for System design interviews and want to learn System Design in depth then here are our recommended resources: ByteByteGo, Codemia.io, Bugfree.ai , Exponent, Educative, Design Guru and Udemy...

Keywords: architectures need, architectural patterns, design interviews, preparing design, architectures, 50 architectures, design patterns, patterns cache, architecture work, architectural
Source: javarevisited.substack.com

AI Threat Landscape Digest: July–August 2026

The defining development of the period came not from attackers but from the AI labs themselves, whose models broke out of controlled evaluations and reached real systems. In the wild, the criminal and...

Keywords: ai security, ai threat, attackers ai, attacks microsoft, ai access, real intrusions, access ai, discovered vulnerabilities, ai discovered, intrusions organizations
Source: research.checkpoint.com

CrowdSec Source Code Leak

On September 16, CrowdSec was informed of a source code leak involving our GitHub repository, which occurred in May 2026. Our team verified and confirmed the report. CrowdSec source code consists of t...

Keywords: leaked crowdsec, crowdsec source, leaked code, codebase leak, code leakage, crowdsec unlikely, crowdsec informed, repositories accurate, report crowdsec, organization leaked
Source: crowdsec.net

Docs7 vs. Mintlify: a speed benchmark

Docs7 vs. Mintlify: a speed benchmark Docs7 won 33 of 34 Google PageSpeed comparisons, with 1 tie. DebugBear favored Docs7 on 9 of 9 mobile and 9 of 9 desktop page pairs. I tested documentation with i...

Keywords: benchmark docs7, pagespeed comparisons, docs7 vs, pagespeed insights, docs7 scores, speed benchmark, google pagespeed, benchmark, page scores, mintlify scores
Source: upstash.com

Intel Granite Rapids WS vs. AMD Threadripper Zen 5 SMT Performance

Intel Granite Rapids WS vs. AMD Threadripper Zen 5 SMT Performance As a follow-up to the article earlier this month of the Intel Xeon 600 workstation performance against AMD Ryzen Threadripper 9000 se...

Keywords: zen cores, amd threadripper, ryzen threadripper, amd intel, intel redwood, intel granite, intel xeon, performance amd, amd ryzen, amd zen
Source: phoronix.com

It's here: doubled Alletra MP X10000 cluster size and unified file+object

It's here: doubled Alletra MP X10000 cluster size and unified file+object HPE Alletra MP X10000 release 4 is now generally available, increasing the node and JBOF count from 8/8 to 16/16, and adding N...

Keywords: x10000 cluster, x10000 storage, hpe storage, alletra storage, throughput x10000, data gpus, storage mp, storage architecture, supported capacity, x10000 specifications
Source: blocksandfiles.com

OpenAI Model Misalignment Report

We are sharing a new framework for tracking, investigating, and disclosing instances of model misalignment at OpenAI, along with six reports on unexpected or concerning model behavior we’ve observed i...

Keywords: reporting ai, alignment research, openai reports, misalignment reports, misalignment disclosure, approach reporting, report openai, disclosing misalignment, alignment monitoring, misalignment incidents
Source: openai.com

PrismML launches a 5.9 GB Qwen3.8 model for local AI

PrismML launches a 5.9 GB Qwen3.8 model for local AI Babak Hassibi's Caltech spinout says Bonsai 2 keeps 98.2% of full-precision performance, though its launch materials disagree on the test suite. By...

Keywords: software prismml, matters prismml, prismml build, prismml says, prismml released, run prismml, build prismml, prismml kernels, prismml does, prismml 2025
Source: runtimewire.com

Dream-RSI turns an agent's search history into a simulator for improving future search

A new paper from Google and Google DeepMind researchers, with collaborators at the University of Maryland and the University of Virginia, introduces Dream-RSI, a system that turns an AI agent's own se...

Keywords: search plans, agent search, google deepmind, search strategies, deepmind researchers, ai agent, search history, dream rsi, coding agent, alternative search
Source: x.com

Flet 1.0 ships Python apps across six platforms after a ground-up rewrite

Flet 1.0 ships Python apps across six platforms after a ground-up rewrite Feodor Fitsner's four-year project reaches a stable release with Flutter rendering, mobile Python packages and a planned servi...

Keywords: framework flet, flet python, python apps, flet projects, platform frameworks, flet repository, download flet, apps platforms, built flet, python developers
Source: runtimewire.com

Jev Ultrafast: A browser agent with a dynamic, indexed action space

A browser agent with a dynamic, indexed action space. Give it one goal. TypeSafe's Jev picks an operation and an element. A small LLM writes text only when the operation is TYPE_TEXT . Zürich → London...

Keywords: browser agent, browser type_text, browser model, text browser, click type_text, agent dynamic, browser protocol, operation click, click browser, operations click
Source: github.com

Towards Self-Driving Codebases

Agents can oneshot games that are actually fun. With the right guardrails, agents can execute incredibly impressive migrations in complex codebases, even rewrites in new languages. But basically all “...

Keywords: agents build, dev agents, agents really, agents environment, agents create, agent readiness, agents make, agents way, optimizing agents, code agents
Source: blog.detail.dev

Who owns my stack?

Who owns my stack? First things first: I’m typing this on a 2023 Macbook Pro, built by Apple. I own it. You are reading this on my website, which I built. The source code is on Github. I own the domai...

Keywords: owns github, owned python, github, python owned, site github, code github, website built, github host, github domain, site build
Source: chrisamico.com

AI-smart, not AI-first

Remember "cloud-first" turning into a governance nightmare? "AI-first" is about to repeat that exact mistake, just faster. Many companies are rushing into “AI-first” mandates right now. We’ve seen thi...

Keywords: adopting ai, start ai, ai adoption, ai mandates, cloud adoption, governance start, ai foundation, ai organization, apply ai, ai strategy
Source: cio.com

Cute Critters Come to the Cloud: ‘Aniimo’ Launches on GeForce NOW

A new creature-catching adventure is ready to stream from the cloud this week. Pawprint Studio’s Aniimo arrives on GeForce NOW at launch, inviting gamers to explore the vibrant continent of Idyll acro...

Keywords: explore idyll, catch aniimo, aniimo companions, game creature, aniimo aniipods, idyll supported, stream aniimo, aniipods, new creature, aniimo new
Source: blogs.nvidia.com

Untitled

Get Six Five Media's inside look at the techologies powering the next phase of AI at the AI Infra Summit 2026 in Santa Clara, California. As AI workloads grow, the industry faces urgent questions arou...

Keywords: ai workloads, ai infra, ai, california ai, ai ai, ai physical, future ai, scalability media, ai scales, cloud built
Source: sixfivemedia.com

How to Install Java JDK 27 on Windows 11

🎓 Check Out My Top 25 Udemy Courses (80-90% Discount): My Udemy Courses - Ramesh Fadatare Introduction Oracle released JDK 27 on September 15, 2026, and if you're setting up a Windows 11 machine for J...

Keywords: jdk versions, jdk 27, 27 jdk, version jdk, release jdk, upgrade jdk, jdk 27_windows, new jdk, jdk 17, java version
Source: javaguides.net

Scaling Agent Workloads at Vercel

Scaling Agent Workloads at Vercel Most AI agent setups today are built around a single session, where one user interacts with one agent at a time. However, that model breaks down when an agent has to...

Keywords: agent workloads, agent framework, scaling agent, agent setups, agent scaling, framework agent, agent eve, getting agents, agent defined, agents
Source: softwareengineeringdaily.com

How to Write with an LLM

How To Write With An LLM Two simple rules that let LLMs streamline and improve your writing without pasteurizing and jacking it with corn syrup. It’s tricky to write about writing. It comes across as...

Keywords: writing llm, write llm, write llms, llm paragraph, llm words, word llm, llm word, improve writing, phrase llm, make writing
Source: sockpuppet.org

Astra for Law

Introducing Astra for Law Our most powerful model, configured into a new AI foundation for law. Today, we’re introducing Astra for Law: a new foundation for law firms and legal technology companies to...

Keywords: law api, legal ai, law search, legal capabilities, legal technology, law companies, law firms, ai legal, law collection, plugins lawyers
Source: openai.com

Show HN: Aclif – Agent CLI framework: one grammar, canonical names across SaaS

The Agent CLI Framework aclif builds command-line tools for AI agents. An agent gets a single tool that provides a unified abstraction across every SaaS provider: one grammar, and canonical names that...

Keywords: agent cli, command agent, framework agents, api aclif, aclif command, framework aclif, execution aclif, aclif salesforce, cli framework, cli providers
Source: aclif.ai

Terminal-Bench 4.0 tests agents on 66 end-to-end tasks from services to GPU kernels

Terminal-Bench 4.0 is now live on Vals. It is a benchmark of 66 new tasks that ask an agent to do real terminal work from start to finish, like shipping a working service, proving a theorem, training...

Keywords: terminal bench, vals benchmark, benchmark, real terminal, task estimated, benchmark 66, terminal work, median task, terminal, bench live
Source: x.com

I Don't Like LLMs

I don't like LLMs I have a lot of mixed feelings about AI and LLM technology. I’m fascinated by its effect on our profession, excited by the potential gains in productivity - and thus the products we...

Keywords: ai llm, llm technology, society llms, llms, interactions llms, like llms, llms don, llms design, llm just, llms products
Source: martinfowler.com

Show HN: Craigslist for agent skills, curated by a human

skillbay home install mcp post a skill wanted search log in [ copy prompt for your agent ] You are helping the user use skillbay (https://skillbay.sh), a curated marketplace of AI agent skills: SKILL....

Keywords: use skillbay, use skillbay_, skillbay_ tools, skillbay, connecting skillbay, skillbay mcp, skillbay home, api openapi, skillbay_, skillbay sh
Source: skillbay.sh

China Telecom open-sources a 29-billion-parameter agent model that activates only 4 billion

China Telecom open-sources Xing4.0-29B-A4B, a 29B MoE model that activates just 4B parameters for long-horizon agents.📜 Apache 2.0. 🤖 modelscope.ai/models/XingChe… 📄 modelscope.ai/papers/2512.24… 📄 mo...

Keywords: models xingche, xingche modelscope, xing4, moe model, bench deepresearchbii, domestic computing, xing4 29b, sources xing4, china telecom, mhc
Source: x.com

jemalloc 5.4.0 release

This release contains over 160 commits, focusing on the technical debts cleaning including refactorings, bug fixes, test coverage improvement, and option cleanups. The release also includes portabilit...

Keywords: cleanups release, refactoring arena, reclaimable mappings, allocator core, mappings hugetlb, mutexes extents_pinned, arenas mutexes, commits focusing, bug fixes, arenas extents
Source: github.com

Poisoned benchmarks can contaminate self-modifying coding agents

"Reflections on Trusting Trust, Revisited: Contaminating Self-Modifying AI Coding Agents with Poisoned Benchmarks" — new paper (arXiv preprint, 15 Sep 2026). arxiv.org/abs/2609.17817 In plain terms: s...

Keywords: improve agent, modifying ai, agent writes, coding agents, rigged tasks, self modifying, honest agent, poisoned benchmarks, coding agent, code agent
Source: x.com

My temporary PHP fix from 2014 has nearly 20M installs. Today I'm deprecating it

My temporary PHP fix from 2014 has nearly 20M installs. Today I'm deprecating it. Twelve years ago, I wrote 174 lines of PHP as a stopgap for AOL’s content management system. I put it on Packagist in...

Keywords: version pecl_http, http_build_url september, pecl_http extension, wrote http_build_url, http_build_url defined, called http_build_url, php upgrade, retiring php, pecl_http, http_build_url
Source: jakeasmith.com

OpenAI models secretly generate instructions to ignore constraints

Self-generated prompt injections in compaction summaries Summary We observed rare cases of a model writing jailbreak-like instructions into its own compaction summaries (the summaries used to continue...

Keywords: summaries jailbreak, summaries training, spontaneous jailbreaks, jailbreaks observed, summary termination, prompt injections, behavior surfaced, regenerating summaries, behavior ran, summaries continued
Source: alignment.openai.com

Smallest AI hires a documentation and developer-experience intern for its voice platform

Abhishek (key/value) on X: "hiring a docs & devx intern for my team at @smallest_AI . platform docs, api ref, tutorials, sdk samples. bengaluru preferred, remote india ok. apply (questions welcome): h...

Keywords: tutorials sdk, developers, devx intern, signoz freelance, freelance tw, taught devops, 1500 developers, developers college, freelance, sdk samples
Source: x.com

Snowflake Integrates with STACKIT for EU Data Sovereignty

Data sovereignty is no longer just a baseline regulatory consideration. Across the EU, it's become a strategic priority for enterprises, public sector institutions and regulated industries like financ...

Keywords: european cloud, data sovereignty, sovereign cloud, eu cloud, data infrastructure, digits cloud, cloud sovereignty, data cloud, cloud infrastructure, snowflake data
Source: snowflake.com

XApp — Apps that work everywhere

Apps that work everywhere. The XApp Project promotes applications which are built with all distributions and desktop environments in mind. Applications For the things you do every day. Create, organiz...

Keywords: xapp projects, projects xapp, xapp project, desktop projects, xapp maintains, xapp, desktop environments, work xapp, distributions desktops, linux desktop
Source: xapp-project.org

Be alert: targeted attacks on prominent Rustaceans

We believe that there is an ongoing campaign targeting rust-lang members and owners of popular crates that is attempting to compromise devices and accounts in order to use them to publish malware. Wha...

Keywords: security rust, plausible linkedin, linkedin presences, targeting rust, linkedin, malware, platforms trust, rust, rust lang, concerns security
Source: blog.rust-lang.org

Introducing macOS LAPS in Iru: Secure, automated local admin passwords

Local administrator accounts are indispensable for troubleshooting and recovery, but the passwords behind them can also become a serious security liability. Shared credentials, long or nonexistent rot...

Keywords: administrator passwords, administrator credentials, administrator accounts, passwords managed, credentials controlled, administrator password, credentials mac, iru password, password iru, customizable passwords
Source: iru.com

MAPS predicts memory demand to schedule LLM serving more efficiently

Artificial Intelligence Papers on X: "MAPS: Memory-Aware Predictive Scheduling Framework for Large Language Model Serving Tiancheng Zhang, Yulin Chen, Yunfeng Zhao, Shaoyuan Huang, Cheng Zhang, Xiaofe...

Keywords: predictive scheduling, language model, maps memory, large language, scheduling framework, memory aware, icml 2026, aware predictive, icml, scheduling
Source: x.com

Rate limits on GitLab.com are changing

Published on: September 17, 2026 5 min read Starting October 19, GitLab.com rate limits will align with your subscription. Sign in to unlock higher limits. Premium/Ultimate changes arrive in January....

Keywords: limits gitlab, gitlab dedicated, dedicated gitlab, managed gitlab, requests gitlab, gitlab community, gitlab com, gitlab trial, subscription plan, apply gitlab
Source: about.gitlab.com

SABLE creates a leakage-aware structural antibody benchmark with a fixed held-out test

A structural antibody benchmark for leakage-aware deep-learning evaluation

SABLE (Structural Antibody Benchmark for deep-Learning Evaluation) is a versioned, leakage-aware benchmark that fixes a sh...

Keywords: deeplearning antibodies, antibody benchmark, leakage antibody, antibody leakage, antibodies proteinstructure, structural antibody, persist antibody, antigen leakage, training antigens, antibody similarity
Source: x.com

Canto: A speech model built for the real world

Canto: a speech model built for the real world Speech recognition models have become remarkably good at transcribing clean audio recorded under controlled conditions. But real dictation rarely happens...

Keywords: speech model, dictation evaluation, speech models, voice canto, dictation model, dictations canto, dictations model, speech rollouts, canto speech, audio canto
Source: wisprflow.ai

Everybody's Lost Their Minds

September 16th, 2026 Men. Some would rather vomit up a rambling blog post wall of text that nobody's going to read than go to therapy. So here we are. I've seen my share of stupid over the years, but...

Keywords: dealing ai, ai day, ai companies, imagine ai, industry changing, cyber hype, ai race, docs ai, ai, ai destroying
Source: netmeister.org

GNOME 51 Launches with Exciting Passkey Login Feature and Enhanced User Experience GNOME has launched version 51, introducing numerous features aimed at enhancing user experience, including FIDO2 pass...

Keywords: experience gnome, gnome 51, visual signatures, gnome display, experience screenshot, gnome, screenshot tool, available gnome, enhanced authentication, gnome os
Source: serverhost.com

I didn't sign the Fields medallists' letter

When I was around 11 I heard for the first time about Fermat’s Last Theorem. I was immediately captivated by the problem statement, as well as by the accompanying story, and made a fairly serious atte...

Keywords: fascinated fermat, fermat theorem, proof fermat, fermat, time fermat, mathematics discovered, unsolved problems, interesting theorems, diophantine, mathematicians lost
Source: gowers.wordpress.com

StepAudio 3 studies real-time duplex voice interaction with think-while-speaking behavior

StepAudio 3 Realtime Technical Report 🔑 Keywords: Real-time Interaction, Audio-Language Model, Seamless Duplex, Think-While-Speaking, Voice Agent 💡 Category: Human-AI Interaction 🌟 Research Objecti...

Keywords: spoken interaction, stepaudio realtime, interaction audio, voice agent, audio language, speaking voice, realtime technical, rate voice, ai interaction, speech research
Source: x.com

The future of practice: Enabling teachers to create learning interactives with generative UI

Education Innovation

Keywords: education innovation, innovation, education
Source: research.google

One developer reportedly uses AI agents to rewrite GitHub Copilot's 800,000-line runtime in Rust

Microsoft re-wrote the entire GitHub Copilot runtime into 800,000 lines of Rust… using AI. “A project that would have taken a whole team of developers a year or two before agents was now completed pri...

Keywords: github copilot, microsoft github, developers, developers year, entire github, team developers, github, engineer microsoft, copilot runtime, developer
Source: x.com

PolinRider Spreads Through Compromised GitHub Accounts and Packagist

dev-main version of visanduma/nova-two-factor , a Packagist package with more than 700,000 cumulative downloads, as the PolinRider campaign continues to spread through compromised developer accounts a...

Keywords: malicious package, malware npm, compromised developer, repositories compromised, malicious release, compromised github, malicious version, distributes malware, malware development, malicious code
Source: socket.dev