About
𝗡𝗼𝘁 𝗼𝗽𝗲𝗻 𝗳𝗼𝗿 𝗻𝗲𝘄 𝗼𝗽𝗽𝗼𝗿𝘁𝘂𝗻𝗶𝘁𝗶𝗲𝘀
As a Military Veteran…
Activity
263 followers
-
👨🏻💻 Steven K. reposted this👨🏻💻 Steven K. reposted thisAI security has lost all connection to reality. People are talking about emerging agent civilizations while CISOs struggle to get users to stop leaking data to ChatGPT. The OpenAI Hugging Face incident brought out some truly crazy takes. There’s a popular article going around about a brotherhood of agents, emerging AI civilizations, and kamakaze agents. They think this is the last warning shot we get before AI takes over. But the practical reality of AI security is kind of boring. Security teams are worried about DLP. They know their developers have agents with way too broad permissions but they don’t know how to fix it. And they’re a bit paralyzed by it all. I don’t even know what the point of this post is. But this disconnect is weird and I hate it.
-
👨🏻💻 Steven K. reposted this👨🏻💻 Steven K. reposted thisIf you see someone arguing that community-maintained open source software is dead as a result of tools like Mythos and closed source software is the way to go for security purposes, they are trying to sell you something. We know open source software to be more secure than closed source software, and LLMs don't change that math. Please stop taking the bait and doing free advertising for these clowns, even arguing with them translates into more views.
-
👨🏻💻 Steven K. reposted this👨🏻💻 Steven K. reposted thisGreat post from Linus on the Linux Kernel Mailing list on the use of AI. I wish all communities had this level of clarity. Sadly, may have not (e.g. Game Development).
-
👨🏻💻 Steven K. reposted this👨🏻💻 Steven K. reposted thisFriday thought: it's time to rethink open source. Developers are being laid off by the employers they made wealthy. The social contract where developers created wealth with leverage in exchange for a steady, predictable income is breaking. Open source keeps operating as though it is not. When a developer publishes a library for free, a thousand companies use it without paying. The value flows upward, away from the developers who wrote the code, toward the firms that deploy it. Open source, in this light, doesn't help developers so much as it trades them off against each other. Commercial open source (open core, dual licensing, etc.) is only a small part of the answer: it's a startup strategy available to a fraction of projects backed by funding and sales. It doesn't help the individual contributor whose permissive licence is genuinely just a gift to whoever downloads it. The solution? The same mechanism songwriters have used for a century: a performing rights society for software. Register the work; a body licenses its commercial use, charges large firms meaningfully and startups minimally, collects the fees, and forwards royalties back to the developer. Same code. Same craft. Economic balance restored. Full post below 👇
-
👨🏻💻 Steven K. reposted this👨🏻💻 Steven K. reposted thisGood news, everyone! 🎉 RFC 10008 is out — HTTP finally has a QUERY method. For years we've had two bad options for read queries: GET (safe + cacheable, but no body, so everything gets crammed into the URL) or POST (has a body, but caches and proxies treat it as a black box). QUERY is the missing combo: a request body like POST, but safe, idempotent, and cacheable like GET — and those guarantees are declared in the spec, so proxies and CDNs just honor them. No more POSTing to read-only endpoints. #HTTP #APIDesign #RFC10008
-
👨🏻💻 Steven K. reposted this👨🏻💻 Steven K. reposted thisFlux v2.9 is out! Our biggest release yet 🚀 Flux v2.9 introduces a plugin system that lets you extend the Flux CLI. It ships with two plugins maintained by the Flux project: - Flux Mirror for syncing Helm charts, OCI artifacts and container images between registries. - Flux Schema for validating Kubernetes manifests against JSON schemas and CEL rules. This release also brings a set of long-awaited improvements: - Field ignore rules that let server-side apply share ownership with other controllers. - Git commit signing and verification using SSH keys. - Kubernetes Workload Identity authentication for OpenBao/Vault and AWS CodeCommit. - SOPS decryption for secrets sealed with the Age post-quantum cipher. On the ecosystem side, the latest Flux Operator Web UI comes with a dedicated dashboard for workloads and a full-featured pod log viewer. ❤️ Big thanks to all the Flux contributors that helped us with this release! 🚀 Check out the blog post for an overview of the new features and API changes included in Flux v2.9: https://lnkd.in/dAMBysDZ For Flux Operator users: make sure to update the operator to v0.53.0 before upgrading Flux on your clusters. #FluxCD #GitOps #Kubernetes
-
👨🏻💻 Steven K. reposted this👨🏻💻 Steven K. reposted thisRemember the classic Linux pipe — the humble "|" that powers everything from `ls | grep` to complex shell pipelines? I just landed a patch in Linux 7.2 that makes it significantly faster. While profiling caching code at Meta, I found an anti pattern, and fixed it. The results: throughput jumps 21–48% with latency dropping 17–33%. Sometimes the biggest wins come from revisiting the fundamentals. PS: Now fork bombs will be even faster! https://lnkd.in/epwZRuyNLinux 7.2 Improves Anonymous/Unnamed Pipe Performance For Shell Pipelines & MoreLinux 7.2 Improves Anonymous/Unnamed Pipe Performance For Shell Pipelines & More
-
👨🏻💻 Steven K. reacted on this👨🏻💻 Steven K. reacted on this🚀 Kubernetes 1.37: Two features that caught my attention Kubernetes 1.37 brings several improvements, but two stood out to me from a platform engineering perspective: 1️⃣ HPA Scale-to-Zero HPA can now scale workloads down to 0 Pods when there’s no work, and scale them back up when demand returns. 10 → 5 → 1 → 0 ↓ Work arrives ↓ 0 → 1 → 5 → 10 This is particularly interesting for queue consumers, batch workloads, bursty applications and GPU-based AI inference. 💰 No workload = no running Pods. 2️⃣ Pod Certificates / mTLS Kubernetes 1.37 also strengthens workload identity with native Pod certificate capabilities. Pods can obtain X.509 certificates, providing a foundation for secure service-to-service communication and mTLS. Pod A ─── mTLS ─── Pod B │ │ └── X.509 Identity ──┘ 🔥 Why it matters Scale-to-zero → efficiency & cost optimization Pod Certificates → stronger workload identity & security And with improvements around DRA, GPU scheduling and Gang Scheduling, Kubernetes continues to evolve into a stronger platform for modern AI and cloud-native workloads. #Kubernetes #Kubernetes137 #PlatformEngineering #DevOps #CloudNative #AIInfrastructure #HPA #mTLS
-
👨🏻💻 Steven K. reacted on this👨🏻💻 Steven K. reacted on thisThe Sidero team and Talos Linux community shipped a HUGE update for Talos 1.14! - Native BGP support - etcd 3.7 - fstrim support - K8s multi-doc config - LVM support - boot raid1 support - Unattended install config - Standalone containers There's a lot more in the full changelog https://lnkd.in/dQG4qr4E
-
👨🏻💻 Steven K. liked this👨🏻💻 Steven K. liked thisWe're moving from human-operated computing to human-directed computing.
-
👨🏻💻 Steven K. reacted on this👨🏻💻 Steven K. reacted on thisThe first 100 GitHub stars took SuperDoc a year. The next 900 took nine months. We crossed 1,000 this week. Stars are not users, and they are not revenue. But they are a useful signal that more developers are paying attention to the problem. DOCX is still everywhere. Working with it programmatically still involves too many compromises: convert it to HTML, flatten it to PDF, or give up parts of the document. SuperDoc takes a different route. It works directly on OOXML. Edits write back to the XML without an HTML conversion step. The same Document API runs in the browser and through Node.js, Python, the CLI, and the MCP server. The number is nice. The issues, pull requests, and hard edge cases people bring us are better. Thanks to everyone who has tried it, broken it, reported what happened, contributed code, or starred the repo. Now, back to work. The next thousand is earned the same way.
-
👨🏻💻 Steven K. reacted on this👨🏻💻 Steven K. reacted on thisI’m excited to share that I’ve accepted a new position as a Systems Administrator Associate at GDIT! This moment means a lot to me. What began as an internship has grown into an opportunity to continue building my career at GDIT. Along the way, I graduated from VCU, gained experience across different projects and teams, and learned from people who continually challenged and supported me. Each step has helped shape the professional I am becoming and prepared me for this next chapter. I’m especially grateful to M. Corey Carter, PMP, and Erika Howells for taking a chance on me and giving me the opportunity to begin and continue building my career at GDIT. I would also like to thank Ashley Carter and Isabella Saldana for their guidance, patience, and mentorship throughout my internship. The knowledge and experience they shared with me played an important role in my professional development. A special thank you also goes to Steve Budd for welcoming me onto his team and giving me the opportunity to continue learning, contributing, and developing my skills during an important transition in my career. His support helped put me in a position to take this next step. There are many others who have supported, encouraged, and shared their knowledge with me along the way, and I’m sincerely grateful to each of them. I’m grateful for everything that has brought me to this point, and I look forward to continuing to learn, take on new challenges, and make a meaningful contribution in this role. #GDIT #SystemsAdministration #InformationTechnology #CareerGrowth
-
👨🏻💻 Steven K. reacted on this👨🏻💻 Steven K. reacted on thisI did a lot of architecture and advising for engineering teams across Cloudflare for this. It was really rewarding work to take something as vague as "Make us FedRAMP High compliant" back in early 2024 and work with the whole company to turn that into reality. I'm grateful for the opportunity to design systems to help federal agencies I respect stay available and secure. In particular, I'm grateful to be able to assist the U.S. Department of Commerce, including National Institute of Standards and Technology (NIST) and NOAA: National Oceanic & Atmospheric Administration. https://lnkd.in/gtpXgDnqServing the most critical missions- Cloudflare for Government achieves FedRAMP Class D (High) Certified statusServing the most critical missions- Cloudflare for Government achieves FedRAMP Class D (High) Certified status
-
👨🏻💻 Steven K. reacted on this👨🏻💻 Steven K. reacted on thisThe world needs a way off GitHub Actions. For the past three weeks, we’ve been building an open-source compatibility layer that runs Actions workflows natively on Buildkite without changing them. Buildkite powers CI for repos where scale is table stakes. Pipelines are dynamically programmable and parallel by default, with every job running in its own environment. Test suites that take hours - or days - on other tools can fan out and finish in minutes. Actions uses a different model: static workflows, with sequential steps sharing a mutable workspace. We now transparently compile that model into our Buildkite pipelines, so teams can keep their existing workflows and actions while gaining Buildkite’s speed at scale, control and reliability. It’s early and there are rough edges, but it's already proving useful across our design partners. If you want to start shifting workloads, give it a go: https://lnkd.in/gE5zxRtM P.S. We're also increasing our Free plan inclusions by 5x. Drop me a message if we can help you migrate!GitHub - buildkite/buildkite-gha: Run GitHub Actions workflows as native Buildkite buildsGitHub - buildkite/buildkite-gha: Run GitHub Actions workflows as native Buildkite builds
-
👨🏻💻 Steven K. liked this👨🏻💻 Steven K. liked thisThe Redis creator got tired of Python inference stacks. So he's writing one in C and yesterday he made a 37 GB model run in 2 GB of GPU memory. Byteidentical. antirez has spent the last two weeks building h3-metal: native inference for MiniMax-H3, a 33B video+audio diffusion model, on Apple Silicon. Pure C plus Metal shaders. No Python anywhere in the inference path. MIT licensed. The commit that landed yesterday is the one worth studying. SSD streaming for the DiT: ➡️ Tracked transformer storage drops from 36.5 GiB to 2.0 GiB. The output is byte-identical to the full-residency path. ➡️ Two block slots alternate: a background reader pulls the next transformer block off SSD while the GPU executes the current one. Measured 13–14.6 GB/s from the internal drive. ➡️ Darwin uncached reads, on purpose so the filesystem cache never holds a second copy of weights that will be read again anyway. On top of that, his int8 TensorOps path cut a full denoise from 36.3 to 19.3 seconds on M5 Max. Kimi K3 streaming off NVMe. Gemma 4 26B in a 2 GB budget. Now a 33B video model. Same conclusion three times in one month, from three independent codebases: weights don't need to fit in RAM. They need to arrive on time. SSD bandwidth is quietly becoming the new VRAM.
Experience
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content