📰 Latest Articles & Guides

Strata walkthrough is live: why DeepSeek vanished and stars doubled anyway

Strata walkthrough is live: why DeepSeek vanished and stars doubled anyway
I spent last weekend staring at GitHub star graphs instead of sleeping. When DeepSeek's repo went quiet, the assumption was that the ecosystem around it would collapse too. But our Strata repo hit 45k stars in the same 48-hour window — which made no sense until I dug into the traffic sources and realized the spike had almost nothing to do with the model itself. The video exists because I needed to show that visibility comes from what you build around a model, not from the model staying online. I walk through the exact moment the graph went vertical — 2 AM UTC, right after a Hacker News thread mentioned Strata in the same breath as the disappearing repo. One concrete detail: the walkthrough includes a screen recording of my terminal running gh api against the traffic endpoint every 30 seconds during the spike. You can watch the clone count jump 400% in real time, sourced entirely from developers searching for alternatives, not from anyone following DeepSeek specifically. What i...

No-Bake Coconut-Macadamia Nut Truffles

No-Bake Coconut-Macadamia Nut Truffles
Coconut-Macadamia Nut Truffles Cook: 45 mins | Total: 2 hrs 30 mins | Serves: 30 to 32 truffles You know those moments when you need to bring something show-stopping to a potluck but your oven is already full of other dishes? Coconut-Macadamia Nut Truffles are the answer you've been waiting for. These little bites look like they came from a fancy chocolatier, yet they come together on the stovetop in about 45 minutes with ingredients you probably already have in your pantry. What makes these Coconut-Macadamia Nut Truffles so special is the contrast inside each one: a creamy, sweet coconut filling made with sweetened condensed milk and a touch of vanilla, wrapped around a single roasted salted macadamia nut, then dipped in rich bittersweet chocolate. The salted nut cuts through the sweetness perfectly, giving you that irresistible sweet-and-salty balance in every bite. The sweetened c...

45,530 stars and the model that vanished — why I made the 60-second walkthrough

45,530 stars and the model that vanished — why I made the 60-second walkthrough
A repo I've been following crossed 45,530 GitHub stars this week, and in the same week the exact model build everyone was cloning got pulled from the official release. That combination is why I cut a tight vertical walkthrough instead of another long screen recording — people were asking "is this still the one to run?" faster than I could answer in comments. The video exists to answer one question on screen: what actually happens when you run the current release versus the build that disappeared. I show the star count, the release page, the diff in the config, and the exact point where the old model tag 404s. No narration padding — the whole thing fits in a phone scroll. One concrete detail of how I produced it: I captured the terminal and browser side by side, then trimmed every frame where nothing changed. The final cut is 58 seconds because anything past a minute dies in the feed. The hard part wasn't editing, it was verifying the claim — I re-pulled the release...

Nginx HTTP/2 Pre-Auth RCE Confirmed as CVE-2026-123007

Nginx HTTP/2 Pre-Auth RCE Confirmed as CVE-2026-123007
Five separate security teams stumbled onto the same pre-authentication remote code execution bug in NGINX's HTTP/2 handling. Wiz Research, Ikotas Labs, Metnir, Rapid7, and Xint all found it independently, which tells you how exposed this code path really is. I've covered enough web server vulnerabilities to know that pre-auth RCE is the worst kind, and NGINX is exactly the wrong place for it. The software powers over 225 million websites as the default front door for web architectures, handling HTTPS and routing traffic before a single line of application code runs. F5 confirmed the flaw matches CVE-2026-123007 and has fixes scheduled for the November 18 security release. With HTTP/2 now the standard rather than the exception, this isn't some legacy edge case. Every unpatched instance is one crafted request away from full compromise, and the window between disclosure and active exploitation is already closing. The HTTP/2 Pre-Auth RCE The vulnerability is straigh...

Shipped a MeTube one-click Railway template — here's what running your own yt-dlp actually costs

Shipped a MeTube one-click Railway template — here's what running your own yt-dlp actually costs
I kept seeing people spin up video downloaders on random VPS boxes with no persistence, no TLS, and a download folder that vanished on restart. So I packaged MeTube — the web UI for yt-dlp — as a one-click Railway template: https://github.com/INAPP-Mobile/metube The concrete detail that mattered most was the downloads volume. yt-dlp writes large files, so the template mounts a persistent Railway volume at /downloads and sets YTDLP_OUTPUT to point there. Without that, every container restart wipes your queue history and partial downloads, which is exactly the failure mode that makes self-hosted tools feel broken. The honest cost to the operator: storage. A single 4K video can be several GB, and Railway volumes bill on used gigabytes, so a heavy user can rack up a few dollars a month in disk alone — more than the compute. I put a note in the README about pruning old downloads, because that's the number that surprises people on their first invoice. It's running now for anyon...

Smoky Grilled Tofu Tacos with Tangy Avocado Cashew Cream

Smoky Grilled Tofu Tacos with Tangy Avocado Cashew Cream
Grilled Tofu Tacos with Avocado Cashew Cream Prep: 45 mins | Cook: 5 mins | Total: 50 mins | Serves: ['2', '2 -3'] Grilling tofu transforms it from a plain slab into something with real texture and smoky depth. The high heat of the grill drives out moisture quickly, which means you get those gorgeous char marks without the tofu falling apart. That crispy exterior against the tender interior is what makes these tacos sing. The avocado cashew cream brings everything together with a cool, tangy richness that balances the charred tofu. Cashews blend into a silky base that clings to every bite, while the avocado adds freshness and helps the sauce stay vibrant green instead of turning dull. Together they create a sauce that's substantial enough to stand up to the smoky tofu but light enough that you still taste the fresh vegetables. What makes this combination work is the c...

Java4Shell Exploits Default Configs in Major Java Apps

Java4Shell Exploits Default Configs in Major Java Apps
Three hours and two hundred dollars. That's all it took an AI agent to find a remote code execution vulnerability in OpenJDK, one of the most widely deployed runtime environments on earth. The bug sits in the Base64 decoder and it's severe: a memory-corruption flaw that lets a remote attacker with no credentials overwrite heap memory, read it back Heartbleed-style, and eventually achieve full code execution. We confirmed it works against widely used frameworks like Apache Struts and Spring Boot, running in their default configurations. I've seen plenty of AI-assisted security research, but the economics here are what stick with me. Two hundred dollars in compute is less than what most teams spend on a single cloud instance for an afternoon. When automated discovery gets this cheap and this fast, it changes the math for defenders. There's a wrinkle in how these exploits were demonstrated, though, and it matters more than the announcement admits. How Base64 Pro...

Every document conversion now leaves an audit trail — here's why that changed how agents use it

Every document conversion now leaves an audit trail — here's why that changed how agents use it
When I first shipped the document-to-Markdown converters, the API just took a URL and returned text. No record of who sent what. Then I started seeing agents batch-processing thousands of files in loops, and I couldn't tell a retry from a duplicate. So I added an immutable audit log: every request gets a timestamped ID, input hash, and processing duration stored alongside the response. The surprise wasn't technical — it was behavioral. Agents started treating the conversion ID as a receipt. They'd store it, reference it in downstream prompts, treat the whole pipeline as verifiable. When something failed, they had the exact input that broke it, not just an error code. I packaged the full pipeline with logging at https://x402.freeq.one/tools/document_to_markdown.html The lesson: an audit trail isn't just for compliance. For agents operating autonomously, a verifiable record of what happened, when, and to what input is what makes a tool trustworthy enough to chain into ...

Hy4 Flash walkthrough is up: what 1/20 the cost and half the coding score actually looks like on screen

Hy4 Flash walkthrough is up: what 1/20 the cost and half the coding score actually looks like on screen
I've been quoting eval numbers for Hy4 Flash against Hy4 for a while — 1/20 the token cost, and less than half the coding score on my harness — but numbers in a thread only go so far. People kept asking what that gap means in practice, so I recorded a full walkthrough and put it up at https://www.youtube.com/watch?v=634wiFOHDzo . The video exists because the score gap surprised me too. I ran both models through the same set of coding tasks, same prompts, same temperature, and Flash didn't just lose fewer points — it failed differently. Where Hy4 got stuck in loops on a refactor task, Flash bailed out early with a partial answer, which my harness scored as a fail but which was honestly more useful to read. That distinction is hard to convey in a spreadsheet, so the video shows the two runs side by side, task by task. One concrete detail of how it works: the walkthrough runs each model on identical sandboxed repos, then overlays a diff of the patches they produced, with a time...

GPT-6.2 Astra and Sol: Two Paths to Reasoning

GPT-6.2 Astra and Sol: Two Paths to Reasoning
When a model line splits into Astra and Sol, the choice isn't just about speed. It's about how you want to reason. I've been running both since they launched, and my take is that Astra is worth the premium for professional work, but only if you're actually doing professional work. Sol handles the everyday stuff fine, but it still stumbles on nuanced formatting. The difference shows up in the details. Feed Astra a spreadsheet template from your finance team and it follows the conventions. Feed Sol the same thing and you get something that looks right at a glance but breaks when you audit it. Both models cut token usage on demanding tasks, which helps with usage limits, but Astra's outputs read less like AI and more like a human who actually read the style guide. Here's the gotcha nobody's talking about: "versatile" is doing a lot of heavy lifting in Sol's pitch. It's fast and efficient, yes, but when I pushed it on multi-step reason...