Bought a wireless scanner Brother DS-940DW so I can use it with the phone to quickly scan something. Turns out the scanner can only be charged via computer USB port. Charging brick isn’t enough. Needs that data line to charge. Would love to chat with the engineer that signed off on that decision.

I switch between Linux, Mac and Windows daily multiple times for work and spend way to much time trying to remember to use the right CTRL, SUPER or ALT key for Copy/Paste (among other shortcuts). Any folks doing that with some tips on how to harmonize this or do you just power through?

Can we make @aaronfrancis.com an offer he can't resist to release a linux version of Solo soon(er)? Just saying ...

How a customer reacts to "Your account was locked due to a chargeback" greatly influences what hoops I make them run through. If they apologize/explain, I unlock, ask them to create a new sub. Just costs me the chargeback fee. If they don't, teaching them a lesson is priceless.

A future where your agents are negotiating with a suppliers agents all without human interaction sounds like nightmare until you try getting prices for corporate insurance policies (for a 1 person company mind you) and realize nothing can be done online and you're going to waste hours on the phone.

Today morning started with: rise and shine, ready for a full day of demo video recording. Today evening ended with: damn you Safari and your little quirks that I have to make exceptions for. So tomorrow: Groundhog Day again! (Well actually cycling first… but then recording!)

I'm torn on submitting an LLM created GH issue against an O/S project. On the one and I understand they're getting swamped by automated issues, but at the same time the issue is real enough that I got hit 3 times yesterday alone so it deserves whatever info I can provide event if it's from an LLM 🤷‍♂️

Poor Hippos! I mean Puzzles, sure I get it, they attack, but what's a Hippo ever done... In more serious news: buy the book, mine is on it's way. Deciding what to build in the age where features are just a prompt away needs a good framework.

I've been challenged by the SO to provide a local LLM (ultimately with a Cowork like harness) for use on our private docs etc and was wondering what hardware/model combos are actually using at the moment? Will a top Mac Studio do or no way around 4090 for smth slower but close to frontier perf?

Waiting for the print edition but ebook is available already. In the AI age deciding WHAT to build is even more important than ever.

The „agentic engineering“ founders finding out they were hacked don’t realize they are the lucky ones. It’s the ones being exploited in secret and continuing to do so for long time that are the ones I’d worry about. In an age where data is worth so much, finding a leak = gold.

London, missed ya.

Today in Hamburg: Kebap shop wanted 5% extra charge because I was using a business credit card (N26) to pay. Different card was fine. I mean I don’t like the surcharges for international/business cards either but it’s the first time I’ve seen it passed on to the customer.

Zero Human Businesses… isn’t that just arbitrage while the masses catch up to using AI? Maybe you can make some money in the meantime but won’t every AI builder start trying to build one of those and quickly saturate the market? Won’t personally bet on that one but it’s wild what’s being created!

Almost a decade into a SaaS 99% of the "bugs" reported are documentation issues, user errors etc... But it's the 1% actual bugs that always leave me asking myself: how on earth has no one encountered that case in all this time!? Especially when it doesn't feel like an edge case.

My new end of day routine is to ask Codex to suggest improvements to the test suite, refactoring for maintainability/reusability and reviewing/simplifying code comments. With the test suite I haven't had a refactoring go wrong (so far) and the code comments are also vital (past decisions/fixes).

OMG leandomainsearch.com is gone - I mean it's still there, but @automattic.com seem to have removed the whole word + prefix/suffix functionality or am I just to dumb to find/see it? Does anyone have an alternative?

I've added a special "Copy LLM prompt" to my error messages (when I run in dev mode) which not only has a basic "replicate this in e2e then fix it" prompt but also appends a bunch of technical data, breadcrumbs for my last 10 steps etc. Quite a productivity boost...

Stripe's Managed Payments just went public 🥳, but am I doing the calculation right that as a EU company (with EUR account) charging a customer in USD is going to cost me 8.75% + 25c? 3.25% for the international card + 2% currency conv. to EUR + 3.5% for managed payments Where's my mistake?

Yay for any European alternative, but I don't understand is why we don't build on what's already there. SEPA Instant is free AFAIK and supported across Europe (and chargebacks are left to the courts not a card processor). But let's see: europeanbusinessmagazine.com/business/eur...

I’m spending about as much time getting in Claude Code fixing Openclaw as I am using it. Running it on a VPS is definitely less out-of-the-box awesome than a Mac mini. Using oxylabs proxy definitely unblocked a lot of things. See my fork if you’re interested: github.com/aduggleby/op...

Codex 5.3 is "feeling" and definitely more Zen than me 😆 "There's no rush, so maybe waiting a moment isn't a big deal. I’ll proceed to pull the image on the server now using Docker Compose. It feels good to have a clear plan to follow"

TIL ~/.claude/CLAUDE.MD is lost after a conversation is compacted. So that's why it goes off the walls after it compacts and tries doing dangerous things. There is a fix though using hooks: github.com/anthropics/c...

Will people in the future look back at the introduction of LLMs and think "probably alien tech" just like we do when we look back at the pyramids because we don't understand how it was possible?

Final episode of bootstrapped.fm dropped last week. I launched my company the year @steveofmcleod.com took over and have been a listener ever since. An OG podcast with real talk and problems directly from the trenches. Looking forward to the book! Thanks for inspiring journey so far Steve!

I am so grateful my biz lets me do crazy things like go from a completed foggy Vienna four hours by train to the west of the country to ride my bike for a few hours and then head back for another four. Totally worth it - especially with SAD. And even got support tickets done on the train!

Recently got back into stamp collecting and I’ve realized it’s one of the few areas where it’s quite easy to get your hand on items from a long gone era. This one’s about 150y old - a very different world for the original owner. But I’ve still yet to cross the 2 century mark though…

Spent the weekend finishing my blog article about my cross Europe cycling trip. I hope I’ve added enough details for the next AI model to produce a nice low risk cycle trip for someone else. Enjoy 😉 alexduggleby.com/blog/midlife...

Good old fashioned in-flight shenanigans: www.saxrag.com/tech/reversi... Often overlooked: "SNI leaks all the domains they visit". HTTPS often doesn't stop a MITM/state from knowing you visited beardgroomers.com + Hyperscalers mostly use SNI for SSL. But improvement to SNI are being worked on.

"every Gen Z shares their location with all their friends and their exes and who knows who" Still unbelievable to me, but yes even in Europe 2/3 of "Zoomers" are constantly sharing their location 🤯. It is surprising anxiety is on the rise in the same demo? youtube.com/clip/Ugkxck1...

Season 3 of Hot Money is amazing. If you‘re interested in spies even the slightest this is A+ journalism by the @financialtimes.com‘s Sam Jones: overcast.fm/+AA5WINe-ucs Vienna‘s official intelligence authorities might still be recovering but there’s some real 3rd man still stuff going on.

Apart from a final push to give Codex a go (I've already adopted his AI-working style apparently) but this needs to be repeated more often: "Almost all MCPs really should be clis." -- @steipete.me A tech with decades of security layers (and docs) already exist. 👉 steipete.me/posts/just-t...

Second to last day on the final stretch of my midlifecycle.com from Dublin to Vienna. On the first two parts I was blessed with awesome weather. On the last part, not so much. Weather forecast was completely wrong, but nothing some extra layers can't fix. At least it's dry. Onward!

With all the AI talk, this is the point that isn’t talked about enough. Where is the next generation of seniors going to get their start? I fondly remember the amount of things I learnt from our seniors at my first job. It was invaluable for my career. /ht @thepanel.bsky.social

TIL: "Most Windows on ARM devices that use Snapdragon chipsets don’t support a full boot/install directly from USB" Wanted to give them a try despite expected compatibility issues but this surprised me and I haven't seen mentioned in any review I read ... Bloated cloud only restore says Asus. 🤦‍♂️

Makes me smile to read my little baby is actually helping people get things done... 🥰

One of the few books we each bought as a couple separately 😅 so if you haven’t here’s your sign to do so.

For about a month now I've listened to my Spotify "Liked Songs" in descending chronological order and just kept going. It's like a mini time travel experience, songs from ages ago bring back vivid memories of what I was doing then. Just wished I could see a timestamp. Highly recommend!

Also is anyone else getting a lot of Google Classroom spam? Fake invitations with the spam content in the subject line mostly ... Hope Google fixes that soon. For now any class invites automatically go to Spam.

Is there a way to get ChatGPT to add new files (it has generated) to the Project Files (e.g. Create a summary of all files)? Ideally it could also edit them? Seems like a basic feature, but either I'm lacking in my prompts, or it's just not possible ...

With AI the amount of developers thinking "I can build that in a weekend" is going to explode and this time they actually may be right. I wonder what effect that's going to have on selling to a developer audience?

Not going to lie but I might just need that 😂

Really want them to build something good for the "new/upcoming" AI age of the web. All the arrows are pointing in the right direction, they're assembling a kickass team, but only time will tell.

Two super interesting finds this week on Reddit: rybbit.io - self-hosted analytics, very quick to install on Hetzner and powerful like the old GA, but privacy first. livecallerid.com - finally a spam call identifier for iOS, just installed and hope it works globally as promised.

Harry Otter 🦦 oh wait you said ruin… Then: Winnie the … I’ll leave it at that.

Wise words. If you find a niche and channels that are working, you're not going to blast it out on social media. Mastermind groups, maybe. @robwalling.com said it on "Startups for the Rest of Us" recently: "Build in Public" is "Be Copied in Public" (paraphrased) Now back to work...

Seems I may have summoned the "virtual" Lianna Patch one too many times? Is "Full Patch Mode" trademarked @punchlinecopy.com ? (And yes, it came up with that phrase all by itself...)

Left political ad for Conservative-isch Party. Right bank ad: reads „What will the future bring?“ Makes me wonder if it’s a coincidence or deliberate positioning…

So I'm seeing it work again intermittently. Not going to inform my customers and just wait out my support ticket for now. I would assume Microsoft has phased rollouts and a sudden spike in 401s will cause someone to take a look? But that doesn't explain why over 48h it works now and again.

Called it. Issue returned and now impacting all customers. Well this is going to be fun... This is platform risk and for all the benefits that come from building on a platform there will be days and issues too where you basically can't do much and have to wait for a big company to fix something...