jimduk

Really promising analysis and best of luck with the work. A sincere thank you.

I've got ALS (MND). Completely agree UX is the problem, gazing at a keyboard on a screen designed to stop multiple keys clogging (QWERTY) feels wrong. Some ideas

- gesture based eye movements, maybe two sweeps on a nine by nine grid, which map directly to phonemes

- enormous 4k 75inch tv with thousands of words or ideograms or phrases

- "writing" with your eyes then doing line to text AI to clean up

- standardish keyboard with massive LLM prediction and clean UX for autocomplete/throwaway with branching options

Ideas are cheap so no clue if these work. Also Tobi split between cheap good non-hackable gaming eyetracking and medical products doesn't help. Finally, with ALS you want to communicate about different things and are more tired.

The math for cheap amazon microscopes is often based on the ratio of a digital image. So use a 60" monitor rather than 30" you have an extra 2x magnification. But not better resolution!

Not sure it is a good saying. Brett Devereaux covers it in detail in his "fremen mirage" articles which I find a good read.

https://acoup.blog/2020/01/17/collections-the-fremen-mirage-...

Did this in a shortlived startup 20 yrs ago. Other cofounder was the ideas guy. Pitch was to roll up granular, factual project achievements up into reporting data , and cascade objectives down. This avoids the red/amber/green fictional layer between PMs and sponsors.

Learned two interesting things

- if your tool mandates a philosophy or process, you massively shrink your market

- real pms , sponsors and engineer s buffer their risk by selectively disclosing information. They don't want a permanent record of open, granular outcome information unless they are in a very mature company.

Agree with this. The Lib Dems had an existential crisis at that time (2010) as Labour and the Conservatives were both fighting for the centre ground and it seemed that would become the norm (deeply ironic given Corbyn, Brexit and Truss later).

So Clegg rolled the dice on getting proportional representation rather than first past the post to solve this and lost. He also had a go at being less of an opposition party (where you oppose) and more of a coalition party like in Europe to get things done (post the financial crisis). This failed.

Everyone hates him for tuition fees, but apparently George Osborne told him to vote against the huge raise. Had this happened I doubt people would remember him with the malice they do.

Finally, he always seemed an ethical person (Lib Dems usually are excepting local election shenanigans), so the charges seem odd and out of character, but so did the move to facebook.

I wrote a little while back about narrative control as being the thing which divides the ruling classes from the consumer classes

Could you post a link? Can't find it and am just watching the Queens 70th Jubilee service which I think is pertinent ! thanks

Management is about sensemaking. Management provides the last answer (if one is needed) for 'Why are we doing this?'. It carries accountability for calling 'when to stop', and in the obverse provides belief and is the backstop that 'this can be done'. It therefore also has accountability for changing constraints (money, scope, resource, time) if needed when asked.

Management ties contexts together, so other individuals don't have to ask infinite 'whys', they trust the manager has this covered.

Management can be done by individuals, within the team, or hierarchically. All can work.

Personally strongly disagree with 'managers typically have no skin in the game'. I find it is necessary to care about the outcome and the team and the customer when I work, but YMMV. Do agree that management is a support, though I prefer to use the analogy that it is the 'glue' role

It's always ok to leave a company - if they can't survive without you, you are doing them a disservice.

But it's an interesting question - I've been intrigued throughout my career (5-25 person project management)

"Which 'reality disconnects' are important and need addressing, which can be left to lie, and which ones are fatal?"

Still don't really have the answer but my guidelines would go like this

- There has to be at least one part of the project (hopefully the most important) that is functional and grounded in reality to grow from. If there isn't then scrap the project/ do a new project. ( In consultancy a question to ask a new client was - Tell me about your last project that succeeded and your last project that failed - sometimes they had no successes - that was worth knowing)

- Fight the small battles all the time - How do we know what done is? Are the tests good? Is that code understandable? Are the specs and interfaces understood? Are we shipping sh*t?. You need to keep your principles here or you can't mentor and good people won't follow/ trust you

- Some of the big battles especially the political ones can only be fought in very specific contexts (changing budgets, team structure, new cultures). Put down a marker you'd like to see improvement there, but also state that now is not the time till the fundamentals are done.

- At the end of the day imagine yourself as a new hire in your team. If the new hire thinks 'My boss isn't the right person to lead the team' then think about leaving/changing roles. If they think 'Wow this is a tough job but I can see how we get there' keep on going. If they think 'I respect my boss but no-one can fix this mess' then think about re-negotiating to what can be done(either leave or help re-base things). If the 'junior you' isn't going to stay in your team, then things aren't going to work.

Strongly recommend Rands in Repose - Bits, Features, Truth

Also Managing the Professional Service Firm By David Maister Helps understand how consulting type firms as a whole work

Question - I've managed (bottom-up, servant-leadership, small smart teams) a few very very smart people who may or not have had ADHD, or been partially on a spectrum or ... I truly appreciated the awesome work and have been open and honest and dealt with all sorts of emotional issues and variable progress on simple tasks and diving down rat-holes etc and seen some amazing overall results and had some great co-workers. The area I really struggle with is when people get temporarily obsessed with a topic that is so far removed from the project and its goals, that I can't cover them e.g. they want to completely redevelop an established, adjacent area of expertise (new to them) from scratch. Any advice on how to have this dialog up-front e.g. we have a software project and I'll back you a long way on these areas A B C for good exploratory work, but designing a new compiler or doing a full org redesign or building custom hardware is not something I can back you on so can we agree that at the outset and just trust me, or let's agree a protocol for how to deal with these situations?

Thanks, we are pretty low budget. Image circles are in the 1/3", 2/3" range and we are looking at patterns with say 5 to 50 um pitch on a flat planar object, monochromatic, we control illumination, standard 3.45 um pixels (so 1k to 5kish). We have a decent small design with two elements (good planarity, ok resolving and ok ish depth of field are what we care about). But we also sometimes look at larger frame (aps-c and up) and are also interested in smaller (eg VGA Omnivision type tiny module but with a decent planar macro lens e.g. f= 4 or 6mm, for 0.5 or 1x mag would be interesting). We do imager based metrology at the sub um and nm level

The back to back infinity-focussed objective set up works nicely with 2 sets of cheap infinity microscope objectives 2x or 4x back to back in a thorlabs tube. Paired lenses cancel out a lot of aberrations (but not all of them). I use this setup regularly for fairly close up, highly planar projective work. For 'low magnifications' say 2:1 and up the Edmunds 2X, 0.15 NA, Ultra Compact Objective has decent price/performance, but this is for sensors in the 1/1.8 - 1" range. If you come up with a nice, planar, cheapish 1:1 design for large sensors, drop me a line. (Disclaimer - not an optics person, but peered into the fascinating rabbit hole)

Agreed, and I would add pre-web well authored Microsoft help files/ hypertext were often really valuable and could be very well done. Also you could build them after the software was done, or before, as the spec in progress. Think msft help was one of the unsung heroes of their success. But if you have changing software Google/SO is much faster. (Disclaimer, used to work on 'performance support' systems late 90s, one of the points of focus was how do you help users keep improving their combined tool use, team performance and business understanding)

Was working with the x264 crowd fun or exasperating or ...? Video-encoding is such a polymath topic (compression, perception, optimisation) and that code was so smart. I clearly remember mb-tree coming out and x264 quality/bitrate performance changing from good to supergood, and reading Dark Shikari's check in note about taking the implicit dependency graph and using it to propagate the relative quality of the blocks and it was a 'wow' moment. Also the Loren Merritt quote page still holds up well http://www.x264.nl/developers/Dark_Shikari/loren.html . Thank you

Weirdly, IMHO, one of the great advantages of the early BBC website infrastructure was directly because they had no state or need for customisation or dynamic pages.

Some BBC Techs told me, back in the early days (1997?-2004?) when other companies were fighting with Vignette and Bea and IBM etc. and worrying how to balance the database, thread, webserver pool for high scale sites, and falling over at every high traffic moment, the BBC had this monster static Apache estate off loads of servers, with a hand-built Perl scheduler which essentially copied your 'subsite' or changes to a subset of these static servers, from whatever static ingest you gave it. This meant as long as you had a semi-centralised site/ information architecture and a good team on the Perl scheduler and load balancing; any internal content subteam could go off and build what they wanted, with whatever CMS/hand tools they had, as long as it spat out a static staging site the Perl scheduler could handle. This model/process/ org structure way out-performed the rest of the UK internet in uptime/performance/cost, and it let a thousand flowers bloom; but only for a one-way read only system - no custom content/ ads/ offers/ personalisation - probably hard to even do centralised log analysis.

<NB agree with Lattelazy - do let the BBC keep innovating, they have come up with some great stuff in content and tech, so keep keep giving them the chance>

Hi, apart from Hayek (and maybe Ursula Le Guin, and A Hamilton), has anyone got any good pointers on theories for when you favour local over global knowledge or vice versa. Either political theories, or game-theoretic. Transaction cost economics is the closest thing I've seen, and maybe some of Stuart Kauffman's stuff on the adjacent possible and the CEO's job being to 'recut' the company into different divisions. But I've never seen anything along the lines of - if I have a changing fitness landscape L and a set of communicating agents A with different sets of knowledge, and they can use their knowledge to extract value from the local landscape, or they can make tools to increase the value of that knowledge or they can communicate or they can listen, then in what situations should you take different actions. The extremes are easy ( local changing landscape, high comms costs, then don't listen to anyone else; static similar landscape, cheap comms, listen to the cleverest agent); Maybe there is some biology stuff with ants or some agent-based modelling work ?

Thanks. So for the manufacturing class, do you mean things like creaform or similar tools from hexagon and atos and the like, or are there some specific niches for 'slower but much more accurate ' photogrammetry ?

Question - if you had a really accurate fiducial (say 1/1000 or 1/10000 of a pixel absolute accuracy, sub micrometre) that could be fixed on/ near the model - is that interesting and would it help speed up the photogrammetry ? We have a system for accurate measurement x,y,z,rot by imaging a flat scale and are currently focussed on precision engineering/ microscopy/ xy stages; but I didn't realise the big photogrammetry systems were so slow or desired micrometre accuracy. May be a whole set of questions about depth of field and mechanical and thermal stability, but just a thought

Dumb question - if I want to do simple image processing on a pi4 (2d ffts, small kernels, summing 2d arrays in one dimension, finding Maxima), and I care about performance, is this a reasonable stack to use,with decent prospects or is it faster/safer to stick on the Arm, despite the GPU. 1k x1k monochrome images, at 3-10 fps ( or more)? Jetson nano seems to be the obvious commodity but pricier alternative with GPU access, but smaller ecosystem.

Agreed. It might (or might not) be lovely if personalities and culture and power and ways of learning didn't come into things, and everything was truly meritocratic/anarchist/optimised or whatever, but this is not the human condition, and never has been.

Much better to understand the working reality and either live with it or try to change it from a position of understanding.

This is what culture and history teach, whether its Xenophon or Succession.

Slightly off-topic generalisation, but maybe one of the reasons hackers disdain good management consultants, is that the latter are experts at understanding "the bits that are unspoken about the operating nature of a company", whereas the former see this as unnecessary overhead to doing what they want to do.

Hikvision are an interesting company with performant products. You can view CCTV as having had 3 stages. Stage 1 - the model is like a TV studio - closed circuit TV, analog cables, control rooms, monitor banks (some setups are still like this). Stage 2 - cameras are network devices - Axis the big European security camera co started life as a networked printer company, not a coincidence. Stage 3 - connectivity is all IP, but the focus is the combined sensing effect of everything on the network and the smarts to make sense of it all. Some projects have tried to do this as middleware, but much easier if you provide the whole solution like Hik. Stage 4 is combined 'pattern of life' analysis from external and on-person surveillance, maybe this comes from Google/Apple, maybe from Hik/Huawei etc. The game is afoot

Another take on conviviality is from its root - 'to live with'. Hannah Arendt (and I'm sure many others) argues that 'living with' other people, well, in a polity, where you inevitably have different views, is a key problem of modern life. If you are strong at sciences as a kid, it often takes you a while to understand this.

If you can get the BBC - this radio program (In our time) is a good introduction. https://www.bbc.co.uk/programmes/b08c2ljg

Agreed both being very impressed by Illich, but it not being workable. I dimly recall The Alphabetisation of the Popular Mind arguing that possibly you shouldn't teach language. Maybe this has merit (on the lines of Lanier arguing the midi-fication of music is detrimental which is clearer), but is the analogy that musical notation shouldn't happen.

My Illich-worship hit a bump when talking to an Aussie English Phd who was much cleverer than me. After my first sentence she said "the problem with people who like Ivan Illich is they just want to talk about Illich". Touche

But well worth reading, and a convivial life is a good life.

Support for the Ozzie comment - Groove was an interesting thing - strong, user friendly, permissions model built on groups as the user perceived them (as opposed to the corporate LDAP model). Once had a (cancelled) gig to deploy it for sheep farmers in Albania as the trust model mapped. IMHO Ozzie definitely understood the user perspective.

All of the above, Life and Fate is essential, but also add Middlemarch: George Eliot - grown up problems, don't lie, be careful around money Bleak House: Charles Dickens - avoid lawyers, don't let money get to you

More left field - Killing Time: The Autobiography of Paul Feyerabend - consistency is over-rated; any Houllebecq - keep thinking; Quartered safe out here - George McDonald Fraser - generation close to ours, but war-time experience like soldiers for the last 2000 years.

A build on this, is as a soft-core participant in a game, often it is more enjoyable to participate in a game where the meta is not known, and everyone is in the 'finding different strategies' mode - bit like early days starcraft. Now the Internet exists, everyone jumps to semi-optimal strategies by copying, not putting in their own thinking time. While I admire high level play, at mid/low level it is more fun if the meta is not nailed down, also philosophically I think life is more 'hidden' than people think, and the idea there is a best-move/ common optimising strategy without thinking yourself is detrimental. I personally would like games which are more random, or where optimising is harder and less discoverable and not shareable.

I think at various points in life, you worry about the whole system and its effects on people, not just maximizing your local niche. This seems good, both morally and from a systems perspective.

For me when younger this was about Ivan Illich (deschooling society), Feynman (the Brazilian light polarization story) etc. Now I'm older it's reading the anger in 'Excellent Sheep', watching the pain in low-trust workplaces, and telling my kids I value real learning over test-passing (they listen, but huge pressure to conform).

It seems true that a new field is more intellectually stimulating and liberating than a codified one (I've talked to older female analyst programmers from the 70s/80s who had a blast with computers and left when it got rigid and boring, I worked in Gurgaon early 00s when those interested in IT were self-educating geeks doing it for love, who later got moved aside + down when the money arrived). Maybe this is why IT people chase the new shiny, as a defensive strategy against codification, a red queen game to keep the game interesting ( though inefficient).

Anyway, glad Paul is publishing these viewpoints, even if the actions to be taken are not clear. Also, society is over-geeked currently. It's not obvious that hackers ( or intellectuals or logistics experts or technocrats) should have too much power instead of History students, or politicians or sports-stars or whoever. Life is both a game and not a game, short term gaming has long-term consequences. History and Biology and Literature teach us this clearly.

Computer Vision and Pattern Recognition. Was a good name, still will be. Great progress has been made, but it's not AGI as anyone who goes to CVPR will tell you.

As you point out, the books don't solve consciousness in a formal way (or they would be more famous), and they have a significant amount of analogy. They are more 'grounded' than say Daniel Dennett. I liked the attractor analogy because a) at some level it addresses why we don't ( normally) get stuck in a conscious rut - e.g. we are not like the sphex wasp discussed by Hofstadter who always repeats the same loop b) It resonates with the idea that your conscious mind wanders, but its experience seems linear. However to get insight of scientific value one would have to follow up on the ideas in Edelman's work and see if they were actually neurologically correct. I haven't read either for a while, and I do maths/computer stuff not neurology so have no expert view. I think there's no harm in reading the Remembered Present quickly, it's fairly easy. Edelman worked with Tononi - I haven't followed his work, but the Wikipedia page on Integrated information theory doesn't appeal to me.