The @arduino library for @EspressifSystem ESP32/ESP8266 is at 0.1.2 too: certificate pinning now works on ESP8266 over HTTP and WebSocket, and a pin that can't be applied fails the connection instead of quietly going unverified.
https://t.co/UGfYv0N3CL
nodrix v1.0.2 is out: security fixes for the open-source IoT backend on @Cloudflare.
Stricter invites and API tokens, MCP access re-checked on every call, and revoked users and devices cut off from open connections.
Update with Retry build.
https://t.co/emnbGBM0hU
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.
A new blog post is out on Inovus Blogs!
What if websites are no longer primarily built for people, but for AI agents acting on their behalf? What could that mean for the future of the web?
I've explored this idea in my latest article. Do check it out!
https://t.co/L9HGPzC87d
@halalstocktok And for speed, we don't guess from pings: we actually download & upload real data (streamed to @Cloudflare for ~2s each way) and report the sustained rate after the connection ramps up.
Runs hourly by default (configurable), or on demand with one click.
netmon v1.3.0 is out ๐
Wi-Fi or ISP?
netmon pings your router and the internet separately, so you know who to blame.
New: on-demand speed tests, more accurate bandwidth, Windows Wi-Fi info.
Open source - single binary - no cloud
https://t.co/CM1I9n6QEi
#selfhosted
@halalstocktok 100% agree.
That's why we records continuously, not just on demand: latency, loss, jitter, DNS and Wi-Fi signal every 60s (down to 10s if you want), all timestamped.
When a call drops, you can look back at that moment instead of running a speed test 5 minutes too late.
makleen 0.3.0 is out, now on @MacHomebrew.
Free up space on your Mac, or ask @AnthropicAI Claude Code or @OpenAI Codex to do it. It explains what's taking space, shows a dry run, and cleans only what you approve.
brew install decoded-cipher/tap/makleen
https://t.co/Lc663yVLiY
For context, I'm running WhatsApp for Mac v2.26.41.74 on my M4 MacBook Pro.
The sidebar and overall UI experience have been genuinely frustrating. How did this make it into production? ๐ญ
@WhatsApp, please take a look.
#WhatsApp#macOS#UXDesign#BugReport
What the hell is happening with @WhatsApp for Mac? ๐ญ
The latest UI feels like a downgrade: awkward sidebar, clunky navigation, settings all over the place.
Who asked for this? What problem does it solve? Did anyone test it on a real Mac before shipping?
Am I the only one? ๐
@HashgraphOnline Special thanks to @HashgraphOnline
for the question that found this bug. Protection patterns with any capital letter silently protected nothing.
Fixed in makleen v0.2.1 ๐
https://t.co/Lc663yVLiY
Released makleen 0.2.0: a ~65 KB CLI that frees disk space on a dev Mac.
It finds tool caches (npm, Xcode, Docker, Gradleโฆ), build folders in projects you haven't touched in 60 days, and leftovers of apps you deleted. You tick what goes.
Free, MIT.
https://t.co/Lc663yVLiY
@HashgraphOnline The path was lowercased but the pattern wasn't, so any pattern with a capital letter silently matched nothing.
Fixed in v0.2.1: patterns ignore case, and ~/ paths, trailing slashes and bare folder names work too.
Thanks for flagging it!
https://t.co/pI9mfrYLTX
Every new article now posts itself, with an AI-written intro, a tag for the author and hashtags. All 144 older posts are in too, with their original dates.
Full thread on Bluesky: https://t.co/DbS48MVm2N
Only recently heard of https://t.co/3Py7LGafKg and wanted to try it, so I gave Inovus Labs its own home on @bluesky: https://t.co/5ZZhC3xxTA, on our own PDS running on @Cloudflare Workers.
Our @Ghost blog is now a proper publication there, with a rich card for every post.
Moved my @bluesky and @tangled_org account to my own PDS this weekend. Same handle, same followers, nothing to re-follow. It runs on a @Cloudflare Worker using Cirrus by @ascorbic.
Thanks to @icyphox for introducing me to atproto.
Rebuilt https://t.co/jBtNRCsPPz and kept it boring on purpose.
No heavy framework shipped to your browser, no JS of its own, no junk. Just HTML, 3 KB of CSS, and some nice type. Pages load before you notice.
How it's made: https://t.co/fVe14gVqTN
AI has made building software ridiculously easy. And I love it.
But sometimes, I miss the struggle.
I wrote about learning to code, AI, and the two horses I still remember from my first website.
https://t.co/Aq5iEnx1Yk
Fine-tuned a 7B model over a weekend without writing a training loop โ @UnslothAI Studio on a MacBook.
The lesson: fine-tuning fixes form, not facts.
246 examples taught it my library's voice fast. It never stopped inventing constants.
https://t.co/6w8zBru3F1