mikehall314.bsky.social
Web developer and Doctor Who fan (not in that order). Science, reason, and critical thinking. Skeptical of the Placebo Effect.
Director of Merseyside Skeptics. Co-host of Skeptics with a K. Director of The Skeptic. Director of QED. Autistic. /H(?:e|im)/ui
433 posts
730 followers
136 following
Regular Contributor
Active Commenter
comment in response to
post
There's fewer than 50 tickets left for the final QED! Pick up yours today:
qedcon.org
comment in response to
post
Haha thank you! We covered open label placebo for PMS recently. Episode 417.
comment in response to
post
You might also be interested in the talk I gave at QED a couple of years ago.
youtu.be/FL96to_4NYU?...
comment in response to
post
Thank you! It’s a topic I’ve been covering on my podcast for around ten years, and more recently written several articles for The Skeptic. I’ve another article on Tuttle coming out in a couple of weeks; my conclusions look to align with yours.
comment in response to
post
#antanddec #metrocentre #jimmynail #eldonsquare #fogonthetyne #stjamespark #aufwiedersehenpet #geordieracer #stottiebread #spuggie #angelofthenorth #nahyerreetillhaveacake #bykergrove #weyayeman #tynetees #ahdecmanmeeyes #divventganpaintballinglike
comment in response to
post
I haven’t to be honest, never really occurred to me as being of more than cursory interest?
It might be fun to write up in a spare afternoon.
comment in response to
post
We solved the same problems, we just solved them in ways we wouldn’t have thought of until we were forced to.
In every sense, the end product was better because we had to design to the extremes.
comment in response to
post
It was immensely satisfying watching our product behave itself across such a range of devices and circumstances. All of which came from the initial constraints we had, and our choice to design to those extremes.
At no point did I feel like the product was compromised by the constraints.
comment in response to
post
But we also found it worked well on shit airport WiFi. It worked on a train, while the phone switches from mast to mast as they speed past. It worked on crowded shared connections. It even worked on dial up (we checked!)
comment in response to
post
We also found that it ran on near anything. Lynx. Nintendo DS. PS3. Screen readers. Those webOS browsers you had in televisions. Every phone I tried it on. (I once went into a phone shop in Florida and tried it out on every flip phone device there).
comment in response to
post
The final product was lightening fast, even on the tiniest whisper of a data connection, and on the most low powered devices. Competitive products were still there rendering a spinner and our product was loaded and running.
comment in response to
post
We made great use of TinyPNG for squeezing the most out of our PNG files; it’s a tool I still use to this day.
Everything was minified, including the HTML, and gzipped where supported. We used content hashes and long expiry, but we knew we couldn’t rely on this so an uncached page was still < 128K
comment in response to
post
So we also had to support full SSR. If you were on a JS device, Whizz would detect this and enhance the page with animations and loading tricks. Otherwise, we fell back to old style page replacement.
We used the picture element to serve SVG where supported, falling back to PNG.
comment in response to
post
Whizz implemented just the API surface we needed to do what we wanted. Much of it just smoothed out browser differences.
Opera Mini also meant we couldn’t rely on JS on the front end. Opera Mini gave us ~2s of JavaScript time before the page was served. It was enough to tidy up but little else.
comment in response to
post
The same code, however, had to run on a 4K screen on desktop and still look the business. So we had to be responsive all the way from 240 to 3840 px wide.
The 128K budget ruled out any thick client JS framework, so we had to roll our own. We had a minimal lib we created in house called Whizz.
comment in response to
post
From this we agreed on a set of design constraints. The app had to run on Opera Mini to support the feature phones. We had only 128K page budget, including all images, JS, HTML, CSS. We had to support screens which were 240px wide, as this was common on feature phones.
comment in response to
post
Next to nobody had iOS devices. There was some Android, mostly ICS and Gingerbread or below. But 40% of the market wasn’t on a smart phone at all - it was feature phones. No touch screen, d-pad for navigation.
There was 3G but it was spotty and the connection we could rely on most was EDGE.
comment in response to
post
My experience is the complete reverse of that. Tight constraints actually push you toward better solutions.
The project I am proudest of in my whole career had constraints that many devs would run screaming from. The customer was based in Africa and the country had poor telecoms infra.
comment in response to
post
Sure. But why? When it’s so obviously not the case? And repeatedly so? I can get the first time. “Wow! That guy lied! Oh my god. That’s not what we expected. Well fuck me” etc
Fool me once, shame on you. Fool me half a dozen times? Wtf?
comment in response to
post
Computer vision & HUDs are cool. AI augmented reality is fascinating.
But this is a trojan horse for megacorp to get into *all* your interactions.
Friends don't let friends bring Zuck in a backpack on their adventures.
comment in response to
post
If you enjoy the show, leave us a review, and tell your friends! Plus you can take up a voluntary subscription to the show via the the Skeptics with a K Patreon, from as little as £1 per month, where you'll get an ad-free show, as well as supporting the hosts
www.patreon.com/skepticswithak
comment in response to
post
No and I agree most conversations probably don’t need it, but I strongly believe we should have privacy by default and only chat in the clear where there is a compelling need (rather than everything in the clear and only private where there is a need)
comment in response to
post
The only problem I have with Telegram is that conversations aren’t fully encrypted by default, you have enable it specifically and most people don’t.
comment in response to
post
Signal is what I’ve been using for a few years now, yes.
comment in response to
post
Facebook has no respect for your privacy or data. Delete your Facebook. Delete Instagram. Delete WhatsApp. Delete Threads. Ditch your Oculus.
Facebook will run roughshod over ethics and the law to get what they want. The only way to stop them is to stop using their products.
comment in response to
post
The number belonged to an unrelated individual. In this case, it looks like his number is also published on his website. Presumably that is how it ended up in the training data for the Facebook LLM. The number was also in the WhatsApp database.
comment in response to
post
The bot claims to have just generated a random string of digits which seemingly coincidentally matches a real number. But it also claimed at one stage to have pulled it “from its database”, before denying this, and also claiming the number was fictional.