46 announcements. 1 week. If your feed moved too fast yesterday, here’s your refresher—every product, feature, and update we shipped for #BirthdayWeek 2026: cfl.re/4ziK3ck

Some really nice averages, and example of how far are we with SLMs on local machines. OpenCode with enola and Qwen3.8 27b Splash running on MacBook M4 Pro Max. The job that scans enola, in the loop, every night and fills the findings. What it really does? Points agent to the folder, which is Kanban-like, where agent reads all other found/fixed/disproved tickets, picks it's own number, reads very deliberate prompt, scans the code for bugs, picks one it believes is really impactful, and then tries first to prove it, then to refute it, and only if it is as sure as it can be, it fills it in newly founded bugs with specific format. I tested both OpenCode2 and Pi, and my choice is OpenCode2 - less mess about MCP, especially since this is very local workflow, and OpenCode uses subagents as well in this case. Averages of tokens per second throughput were unthinkable just two months ago in a local setup. And this then waits for another such pass, with a different prompt, to do the fixes. Opus 5.5 comes as a final authority for reviews. After Opus says it is fine, fixes don't get merged yet: seven different scripts are ran to prove that nothing got crazy - what we used to call back at the time performance tests. Only if this proves that everything is still how it should be - or better - PRs are merged. If you didn't try enola, check it out github.com/enola-labs/enola

AEO Explorer, almost ready 🔥 - you can selfhost, BYOK and run 4 LLMs easy to collect answers - then later pair answer insights with your GA4, GSC to get real reasons "where is your brand mentioned" and "what works" aeo-explorer.de/

how the #### people work 8+ years in #figma and position thing manual - that every box has different spacing and all but is aligned left-right, yes, manual... where and when the #### this nonsense stops #webdev

Web Search API is now in beta. Ground your AI responses in live web data with zero data retention through AI Gateway. developers.cloudflare.com/ch…

Cloudflare CLI is now in beta. It unifies Workers, DNS, and security management into a single tool that works great with coding agents. developers.cloudflare.com/ch…

We'll be spending a lot more time trying to understand the outputs of language models. A few thoughts, tips & tricks: Writing. Something I've had success with: Ask your LLM to explain something in ASD-STE100, it's a controlled language specification originally developed for aerospace maintenance documentation. LLMs well-versed in this language and it comes with heavy constraints on clean writing style that I often find a lot more readable. Sometimes I've tried to soften it a bit e.g. ask for "80% of the way to ASD-STE100" because the spec is quite stringent. But even better: Diagrams / images. Instead of writing, ask your LLM to create a diagram. These can be a lot easier to process, parse, and understand. But even better: Web pages. Ask for output "in HTML" to get a beautiful, interactive webpage. LLMs are getting really good at frontend and can create beautiful experiences, animations, etc. But even better: Explainer videos. The output format I am most bullish on is fully custom / bespoke explainer videos generated on any arbitrary topic. Experiment with things like "Create a 3b1b style video explainer on X. Use my ElevenLabs API key for audio narration". (you'd need an API key for the latter or you can ask your LLM to find you decent free alternatives that use your local compute). This is actually starting to work! In summary: - As LLMs get better, they will do more and more of the legwork autonomously, and a lot more of our work will rise up the abstractions into oversight and understanding. - Luckily, LLMs can help here too because as intelligence and code are increasingly abundant, you can ask for large, custom, discardable software artifacts (e.g. web apps, video explainers) that would have never made sense to create before. Push the boundaries here and you'll be surprised.

what should I pick for 60k month of headless Next.js website: - @Cloudflare - @vercel - @awscloud or anything better ?

Day 4 of Cloudflare's Birthday Week is a BIG one. There's so many announcements, the majority are new products or features! 🚀 K2: Kafka-like event-streaming service on the edge 🏋️‍♂️ Managed CloudflareOS 🧑‍💻 Artifacts: Open beta & Git platform competition ⚡️ Workers KV Instant: 2ms cache reads globally 📊 Cloudflare Basin: CF Data Platform goes GA 🔐 Modern cryptographic algorithms in Workers 🔎 AI Search goes GA 🤖 Sovereign AI and the fight for choice That's a lot to get through, I'll summarise each below! 🚀 K2: Kafka-like event-streaming service on the edge This is a wild one. This is not Kafka, but it is Kafka-like in its design, so much so that there is future work to make Kafka Clients work with K2. Cloudflare K2 is a serverless event-streaming service that stores durable, ordered streams backed by R2, running entirely on Cloudflare's edge network. This allows you to decouple producers from consumers. Consumers read events in batches through subscriptions, with support for parallel processing or fan-out. It's available today in public beta for anyone on the Workers Paid plan. The limits are low during beta, and there's no charge during beta either. 🏋️‍♂️ Managed CloudflareOS When CloudflareOS was announced earlier this year, it instantly became incredibly popular. Any person or company could deploy it to Cloudflare with a fews clicks. In a nutshell, it gives everyone an agent workspace within an organization, connected to your systems' data and incredibly secure. Here's what's new in CloudflareOS: - Mount Git repos and work with code - Work across Google Workspace - Export work in the formats your team uses Today, Cloudflare is opening a waitlist for a managed offering of CloudflareOS, with signup available in the links below. 🧑‍💻 Artifacts: Open beta & Git platform competition The moment a lot of you have been waiting for: Artifacts is now in open beta. Everyone can try it out for themselves, available from today! Alongside that announcement, here's what's new with Artifacts: - You can now connect an Artifacts repository to a Worker through Workers Builds - Manage Artifacts directly from Workers via the Artifacts binding - Subscribe to events happening in Artifacts via Event Subscriptions (pushed, cloned etc.) - Data jurisdictions for Artifacts repos - Metrics! See all the metrics about your repos in the dash Finally, there is a competition to build the next Git hosting platform on Artifacts. Two members from the top 3 teams will be flown to Cloudflare Connect in SF to demo + the winners get $25,000 in Cloudflare credits and an invite to the VIP Speakers dinner at Connect. ⚡️ Workers KV Instant: 2ms cache reads globally This is one of those announcements that you read, and you can't help but marvel at the engineering feat. Workers KV has been around since 2018 and is perfect for read-heavy workloads that can be eventually consistent (p99 4.4 seconds across the world), with p99 uncached reads of 290ms. What if you need it to be FAST though? That's where KV Instant comes in, backed by Cloudflare's internal KV store, Quicksilver. How fast, you might ask? p99 reads are 1.6ms (yes, milliseconds) and writes are 250ms (that's globally replicated in that time). If the key is cached, the read is instant, hence the name. The API is the same as Workers KV, with pricing that reflects its intended use as high read, low write caching: $0.20 per million reads (60% cheaper than regular KV) and $0.10 per operation for writes, plus $100 per MB, per month. It's available in private beta, with open beta planned later. 📊 Cloudflare Basin: CF Data Platform goes GA At first, this might sound like a new product - but it's actually a culmination of years of work to make Cloudflare a fully-fledge data platform. Basin brings together three key pieces of the Cloudflare data platform under one umbrella: - Pipelines, allowing you to receive events from Workers, HTTP, or Cloudflare Logpush, transform them with SQL, and write them as Apache Iceberg tables or files in R2. - Basin Catalog, which manages Iceberg metadata and automatically maintains tables to keep them fast and cost-efficient. - Basin SQL, is our serverless, distributed SQL engine for querying Apache Iceberg tables directly on Cloudflare. There's a ton more coming here, such as support for the latest Apache Iceberg spec, advanced adaptive table-maintenance strategies and much more. 🔐 Modern cryptographic algorithms in Workers Cloudflare Workers is adding support for post-quantum-resistant algorithms within Web Crypto, available today. The list includes: - ML-KEM-768 and ML-KEM-1024 for key encapsulation - ML-DSA-44, ML-DSA-65, and ML-DSA-87 for signatures - encapsulateBits(), decapsulateBits(), encapsulateKey(), and decapsulateKey() - getPublicKey() - SubtleCrypto.supports() - JWK import and export for these algorithms There's a great rundown of some of these algorithms and why they are important in the blog! 🔎 AI Search goes GA Cloudflare’s AI Search combines Workers AI, Vectorize, R2, and Browser Run into a fully managed index and retrieval pipeline As part of going GA, there's some enhancements to the product: - Multimodal embedding and retrieval, with caption-based understanding and native image retrieval - AI Search now accepts your text files (Markdown, HTML, CSV, JSON and similar) and PDFs up to 10 MiB, up from 4 MiB - OCR support is available too Alongside GA, pricing has been announced - there's many dimensions to it, and the blog below has all the info on pricing. 🤖 Sovereign AI and the fight for choice Cloudflare is bringing EuroLLM, which supports 35 languages, and Switzerland’s fully open Apertus, trained on more than 1,500 languages, to Workers AI. This matters because governments and essential services may be vulnerable if they depend on one AI provider or lose access to a model. In a nutshell, AI should be for everyone and not belong in the hands of the few. --- Today's announcements are just insane to me, there's SO much there and it's hard to fit all the detail in without this turning into a novel! What're you most exicted by?

Bigger problem than AI, is this kind of people + AI: The Corporate Martyr: The person who acts like completing a basic operational task took a piece of their soul. They want everyone to know how much they "sacrificed" for the project. Performative Productivity: The act of prioritizing the appearance of working hard over actually getting things done. They spend more time presenting what they did than it actually took to do the task. #CorporateLife

what is best way to ping @vercel to get onboarded company that wants to do "re-plaform" to Vercel ? @vercel_dev

Cleaning a oldie laptop to spin that @OmarchyLinux

The biggest problem with AI in development right now? It made people lazier at defining tasks. Before, a bad spec was a vague one-sentence ticket. You knew it was bad. Now, people feed a vague idea to an LLM and hand you a 500-word, beautifully formatted document that says absolutely nothing. AI lowered the cost of generating text to zero, but the cost of clear thinking stayed the same. Now developers have to play detective, reverse-engineering actual business logic from synthetic fluff. Stop mistaking generated text for defined requirements. #ai

That mysterious 1px black line under your <video> — it's not a border. Chased this one across a client site today. An autoplaying video shows a hairline dark line — sometimes bottom, sometimes right. Zoom in: gone. Different monitor: back again. What's actually happening: the browser rounds the video layer to physical pixels separately from the element's CSS box. If the box lands on a fractional width, the mismatch leaves a sub-pixel gap showing whatever sits behind. One of solutions is the CSS fix is three lines: wrapper → border-radius + overflow: hidden video → display: block video → transform: scale(1.005) The video overdraws its box by half a percent, the wrapper clips it, no seam can survive on any edge. Works in Chrome, Safari, Firefox. But my favorite finding is upstream: video codecs encode in 2px blocks (macroblocks are 16px). Export a video at 1559×879 and the encoder pads it — and some decoders render that padding as a dark line inside the frame. No CSS can save you. So: always export video at even pixel dimensions — ideally multiples of 16. One checkbox in After Effects settings kills a whole class of bugs at the source. #webdev #css #frontend #videoencoding

Nice, 115 visitors within the first 10 minutes isn't bad for starters 🚀

Check out the latest article in my newsletter: You Can Measure Everything About Your ChatGPT Ads Except Whether They`re Any Good linkedin.com/pulse/you-can-m… via @LinkedIn

was mostly focused making "SaaS", and its almost done - that I just need to decide on Stripe or Paddle. ~ you throw yourself into the code that you forget anything else, in the end you want to run away and not finish it. #overwork #buildinpublic

when you make tools, even small one - and in the middle of convo AI is following on it own by saying: "Now I'll build a decision-panel prompt (options + full context) and dispatch all four in parallel: codex, a fresh claude, grok, and agy. Writing the prompt: ..." Quite happy I created github.com/apsolut/apsolut-a…

The client never sets PRAGMA foreign_keys=ON — and on Turso/libsql that pragma defaults off per connection, so the schema's onDelete: "cascade" may not fire at all. That's bigger than MCP if true. Let me check what the existing test concludes: RTFM!