← Home About Archive Reading Tweets Also on Micro.blog
  • The loading time is the primary thing that makes Apple New+ unusable. I don’t understand how they could have let this design out. Does it load faster in California?

    β†’ 1:16 PM, Dec 23
  • If you’re using a swiping keyboard, be aware that “dataviz” is very very close to “satanic”.

    β†’ 11:56 AM, Dec 15
  • If you’re going to destroy an $80 billion organization over AI fears, at least have the decency to stick to your principles. 😬

    β†’ 9:59 AM, Nov 21
  • Well. That was an interesting weekend, wasn’t it?

    β†’ 10:20 AM, Nov 20
  • It occurs to me the important thing about the hype cycle isn’t just where we are, but how high and low we expect the peak, valley, & equilibrium to be. Two contrasting expectations:

    Blockchain: high peak, low valley, low equilibrium

    GenAI: high peak, high valley, high equilibrium

    β†’ 8:09 AM, Oct 20
  • It occurs to me the important thing about the hype cycle isn’t just where we are, but how high/low we expect the peak, valley, & equilibrium to be. Two contrasting expectations:

    Blockchain (pictured): high peak, low valley, low equilibrium

    GenAI: high peak, high valley, high equilibrium

    πŸ§‘β€πŸ’»

    β†’ 8:09 AM, Oct 20
  • It occurs to me the important thing about the hype cycle isn’t just where we are, but how high/low we expect the peak, valley, & equilibrium to be. Two contrasting expectations:

    Blockchain (pictured): high peak, low valley, low equilibrium

    GenAI: high peak, high valley, high equilibrium

    πŸ§‘β€πŸ’»

    β†’ 8:05 AM, Oct 20
  • The team at Mindset is looking at the Clean UI5 concept and book, which was discussed a year ago during SAP’s Devtoberfest. Seems like nothing much has happened since then? That’s too bad.

    Meanwhile, Clean ABAP has a fairly robust ongoing community discussion based around it’s Git repository. πŸ§‘β€πŸ’»

    β†’ 12:46 PM, Oct 9
  • Am I starting to read the Wheel of Time again? Yes. Did I forget that these books are approximately a bajillion pages long? Not really, but effectively yes.

    β†’ 12:56 PM, Oct 6
  • If I build a user experience survey in such a way that people who care about user experience have a physical reaction to it, will it influence my results? Yes, SAP BTP Cockpit team. Yes it will. πŸ§‘β€πŸ’»

    β†’ 12:40 PM, Oct 5
  • Listening to @ismh@mastodon.social on the Connected podcast not know what a skyway is sure is something. β€œA tunnel suspended above the ground.” 🀣

    β†’ 1:20 PM, Sep 30
  • DSAG position paper: Artificial intelligence (AI) - dsag.de is an interesting read to get a grasp on what the German user-group is thinking about this stuff. More focused on licensing and commercial models than use-cases, but that is to be expected from DSAG. πŸ§‘β€πŸ’»

    β†’ 11:10 AM, Sep 30
  • Finished reading: Piranesi by Susanna Clarke πŸ“š

    β†’ 8:59 AM, Sep 30
  • My team has been working with CoPilot, and working on building SAP-aware in-line assistants (which can be really great, it turns out). Some of this post on ways in-line assistance can hurt definitely hits home. πŸ§‘β€πŸ’»

    β†’ 8:29 AM, Sep 19
  • My team has been working with CoPilot, and working on building SAP-aware in-line assistants (which can be really great, it turns out). Some of this post on ways in-line assistance can hurt definitely hits home.

    β†’ 8:27 AM, Sep 19
  • Foundation season two is what it should have been from the beginning. It was good, sometimes great, throughout and didn’t disappoint with the conclusion.

    β†’ 3:34 PM, Sep 16
  • Pretty happy the RAG (retrieval augmented generation) acronym is now in widespread use. It accurately describes most enterprise uses of LLMs accurately. Prior to this, people talked a lot about augmenting a model with embeddings, and I think that caused a lot of misunderstandings. πŸ§‘β€πŸ’»

    β†’ 5:19 PM, Sep 6
  • Looks like an anti-caste-discrimination bill has been introduced in California. Seems clear that it’s a problem, and legal protection sounds like a good step.

    My assumption is that this is a major problem in the SAP space, just like the larger tech ecosystem, but I rarely see it discussed. πŸ§‘β€πŸ’»

    β†’ 7:41 PM, Sep 4
  • Science is highly political (both internally and externally), and has been since its inception. I always find it interesting that one of science’s most effective political tactics is to pretend to be apolitical.

    β†’ 8:10 AM, Aug 31
  • Finished reading: Nine Princes in Amber by Roger Zelazny πŸ“š

    Went into this with really low expectations, barely remembering it from almost 30 years ago. Pretty pleasantly surprised.

    β†’ 9:29 PM, Aug 30
  • Why would anyone ever close a dialog box without reading it? And yes, I am sub-tweeting my entire extended family.

    β†’ 1:59 PM, Aug 30
  • Looks like SAP TechEd call for community speakers is out.

    TechEd is only in Bangalore this year and you need to be attending the conference to qualify for one of these talks. If you’re planning to attend, this is a great opportunity to get speaking experience and meet others in the community! πŸ§‘β€πŸ’»

    β†’ 10:47 AM, Aug 28
  • New UI5 metadata-driven-controls introduction is helpful for understanding what these are and how they fit in to the UI5 roadmap. Good that these are finally available in OpenUI5, so are usable in contexts where you don’t have an SAP contract in place. πŸ§‘β€πŸ’»

    β†’ 1:41 PM, Aug 27
  • Before and after.

    β†’ 12:57 PM, Aug 27
  • For a moment I misread as XCode and got unreasonably excited.

    β†’ 2:44 PM, Aug 25
  • Python in Excel looks like it is going to make Excel workbooks into full-fledged Python notebooks. This looks really compelling, even if just for visualizations, and when I think about integrations like PowerBI the gears really start turning. πŸ§‘β€πŸ’»

    β†’ 9:16 AM, Aug 24
  • Finished reading: Fugitive Telemetry by Martha Wells πŸ“š

    β†’ 6:24 PM, Aug 22
  • Finished reading: Network Effect by Martha Wells πŸ“š

    β†’ 4:42 PM, Aug 20
  • Finished reading: Exit Strategy by Martha Wells πŸ“š

    β†’ 4:42 PM, Aug 20
  • Finished reading: Artificial Condition by Martha Wells πŸ“š

    β†’ 7:13 AM, Aug 16
  • Finished reading: All Systems Red by Martha Wells πŸ“š

    Re-read this one on paper this time. Still just as good.

    β†’ 6:03 PM, Aug 13
  • Finished reading: Gideon the Ninth by Tamsyn Muir πŸ“š

    Hilariously weird. Really enjoyed this. The sequel Harrow the Ninth is also good and super weird, but very different.

    β†’ 4:38 PM, Aug 12
  • I have to admit, early this morning I read a summary of the LK-99 fracas to date and dismissed it as a parody. I was on the fence until I got to the Russian catgirl part. Turns out, it wasn’t a parody. 🀣 Or was it?

    β†’ 11:18 AM, Aug 2
  • One secondary goal of the two week vacation has been to piece back together a personal knowledge management system again. Think I’ve got the bones in place with Notion and Zapier, and looking forward to putting it to work. I think it’s been 10 years since I had a real system.

    β†’ 5:44 AM, Jul 29
  • Heh. First time I’ve tried Threads since being in northern Europe for two weeks. Turns out that doesn’t work. 😬

    β†’ 7:14 AM, Jul 28
  • The smoke has arrived in Minneapolis.

    β†’ 1:30 PM, Jun 14
  • Oh, more takes:

    1. I didn’t see much demonstration of gesture controls. My guess is that they are still finalizing the control framework. I want to see a lot more of this.

    2. Interactive widgets are probably going to be great on this thing. But we didn’t see them, I don’t think. Interesting.

    β†’ 6:44 PM, Jun 5
  • Overall take: Vision Pro looks like a really interesting product. It’ll take a couple generations to come into its own, but it looks to me like it’s for real.

    β†’ 6:41 PM, Jun 5
  • Vision Pro takes cont:

    1. I really hope if supports displaying multiple desktops from your closed computer by the time it launches. What was shown was one 4K display.

    2. How are they going to do the fitting? Or is it going to come with all the different goggle attachments?

    β†’ 6:36 PM, Jun 5
  • Vision Pro takes:

    1. Seems a lot like Apple Watch. They don’t really know what it’s for, but they are putting it out there.

    2. Price is the price. Whatever. It’ll get cheaper.

    3. Over the last couple years it has become very common to wear AirPods Pro while talking with people.

    β†’ 6:33 PM, Jun 5
  • Wow. ExpressVPN uses some real dark-side account retention design tactics. Time to try a different VPN service.

    β†’ 6:36 PM, May 11
  • It’s really cool to see truly multi-modal models becoming a thing. This has been a topic of discussion in philosophy for centuries. From Hume’s theory of correlation through triangulation (basically, multi-modal learning). The current time is a fascinating opportunity to see what actually works.

    β†’ 10:50 AM, May 11
  • Finished reading A Wizard of Earthsea by Ursula K. Le Guin πŸ“šHard to believe how much fantasy and sci-fi I’ve read without reading Le Guin. This book is poetic.

    β†’ 12:39 PM, Apr 22
  • Um. Is the iOS squirrel emoji actually a chipmunk? 🐿️

    β†’ 5:16 PM, Apr 20
  • So much of the thinking about LLMs and copyright is inexcusably sloppy. If we think training on copyrighted images violates copyright, we’re going to need to explain why my looking at that image in a web browser doesn’t.

    β†’ 11:32 AM, Apr 17
  • Still really not seeing the point of artifact. The new comments don’t seem to add anything either. πŸ˜•

    β†’ 11:18 AM, Apr 15
  • Looks like people are figuring out that off-chain NFTs just point at a representation of the work (usually a URL) and if whoever controls the URL controls the work. Whoops!

    β†’ 7:50 AM, Apr 14
  • The Chris Best answer to Nilay Patel regarding racist/xenophobic content on Substack should probably go on some kind of worst all-time list. We can debate if it’s a wrong answer, but don’t equivocate. Really too bad that Nilay botched the question though, which he explains in the transcript.

    β†’ 7:02 AM, Apr 14
  • First bike-to-work day! πŸŽ‰

    First flat tire. 😭

    β†’ 5:09 PM, Apr 10
  • The whole GM CarPlay debacle has really thrown a wrench in the buy-a-Bolt plan. At this point, probably just going to stick with the ICE car until small EVs with good phone integration become available.

    β†’ 4:42 PM, Apr 7
  • Back after being gone for a week. Cat is keeping a close eye on me as I work.

    β†’ 3:30 PM, Apr 7
  • What would I have liked to see happen with SVB? Something like making 50-80% of uninsured deposits available immediately. The rest probably would have been available eventually. Some longer-term pain for those with foolish liquidity-management isn’t a bad thing. But I’m #notanexpert and FDIC is.

    β†’ 8:40 AM, Mar 15
  • How did camels get such a bad reputation in the design community? 🀣

    β†’ 9:11 AM, Mar 13
  • Deliberate Practice

    I’m slowly reading 97 Things Every Programmer Should Know. Do Lots of Deliberate Practice really struck a nerve. Parts of it are great. Parts are unfortunately misguided.

    I think the great parts are in identifying the nature and importance of deliberate practice. The difference between practice-orientation and task-orientation is so often lost, and is so clearly articulated here.

    Additionally, the identification of parallels between deliberate practice and learning is so important, and I’m glad to see it mentioned here. I would extend further and mention that learning itself is a practice at which you can gain expertise.

    I personally believe that practicing learning is the most important form of deliberate practice we can engage in. That is, not learning something because we need to know it, but learning something in order to practice learning something. As mentioned in Continuous Learning, it sure would be nice to have Neo’s ability to download information in the Matrix. In real life, how does one get closer to that amazing ability? Practice.

    Two big disagreements from me:

    First, the 10,000 hours of practice to achieve expertise is not important or realistic. Time required varies hugely by person and task.

    The 10,000 hours thing has got to be one of the more damaging memes regarding learning. Clearly the idea was the impress on people that expertise requires a lot of work. 1,000 hours seems a little too achievable, and 100,000 hours is too depressing (you’d be dead before you’d be an expert), so Malcolm Gladwell landed on 10,000 hours because you can only effectively market nice round numbers.

    Second, it’s absolutely best … errr … practice to practice something you are already an expert at. Have you ever met or observed someone who has deep expertise? Have you noticed how they still practice what they became expert in ages ago? They do.

    True, usually this practice is disguised as something else. A job, perhaps. Or more often teaching the subject they have mastered. But even for people who constantly exercise their expertise through one of these activities, you will find that they set time aside for deliberate practice, focused on maintaining and honing their mastery.

    β†’ 3:07 PM, Mar 6
  • This thread is hopefully not incredibly surprising to those who pay any attention to racism these days, but the explanation and references are really interesting. hachyderm.io/@mekkaoke…

    β†’ 4:11 PM, Feb 12
  • Bruce Schneier on hacking the tax code. A good voice to listen to on future security and policy issues, and an ML-realist.

    β†’ 8:14 AM, Feb 10
  • In the back of my mind trying to figure out if someone building Nuzzel for the Fediverse would make people angry. Fedi-people seem to get angry about stuff like that. 😬

    β†’ 5:17 PM, Feb 9
  • I found this interview (paid Stratechery subscription required) with Midjourney founder David Holz incredibly interesting. Some of the product creation insights and predictions around cloud build-out that start 1/3 of the way through are really excellent.

    β†’ 4:30 PM, Jan 30
  • I’m happy to report that the ChatGPT prompt, “Summarize what this code does in simple English” generally produces pretty impressive results. But take it to another level with the prompt, “Are there any bugs in this code?”

    β†’ 5:40 PM, Jan 17
  • Need to write at length about where I see generative models like ChatGPT and Stable Diffusion going in the next few years, especially in enterprise. The next year or so is going to be rough, with lots of misguided and damaging projects, but I think years 3-4 look promising.

    β†’ 9:57 AM, Jan 13
  • I feel like chatGPT is going to be the enterprise blockchain of 2023. It’s a real shame, as unlike blockchain, these types of language models are very useful for many things. But there is so little understanding of how these models actually work among some starting to hype them.

    β†’ 6:19 PM, Jan 9
  • Finished reading: Exhalation by Ted Chiang πŸ“š

    Lots of great ideas and compelling writing. I remain not a fan of the short story genre, but I’d read more of Chiang’s.

    β†’ 5:44 PM, Jan 7
  • What Can We Learn from Barnes & Noble’s Surprising Turnaround?

    I haven’t been in B&N in years, but have re-discovered physical bookstores over the last few months. You can tell if a store is run by people who love books the moment you walk in.

    β†’ 8:55 PM, Jan 5
  • Catching up on some reading backlog when I have the time. Am going to be trying to do more of that this year. Currently Blameless PostMortems and a Just Culture - an old, but important blog from Etsy.

    β†’ 11:55 AM, Jan 5
  • The lack of a shoveling workout in the Workouts app is a major oversight. ❄️

    β†’ 8:42 AM, Jan 4
  • Finished reading this a bit ago: Humankind by Rutger Bregman πŸ“š

    Some serious confirmation bias potential with this one, but a really interesting book.

    β†’ 9:14 PM, Jan 1
  • RSS
  • JSON Feed
  • Micro.blog