haruki zaemon

Amazon Prime Video rearchitected their streaming service from a distributed microservices architecture to a monolith application, resulting in higher scale, resilience, and reduced costs.

Werner Vogels:

My rule of thumb has been that with every order of magnitude of growth you should revisit your architecture, and determine whether it can still support the next order level of growth.

if there are a set of services that always contribute to the response, have the exact same scaling and performance requirements, same security vectors, and most importantly, are managed by a single team, it is a worthwhile effort to see if combining them simplifies your architecture.

1. #software-engineering

(via Michael Neale)

A leaked internal document says Google are in a pickle when it comes to AI, and are being “lapped” by public efforts:

Open-source models are faster, more customizable, more private, and pound-for-pound more capable. They are doing things with $100 and 13B params that we struggle with at $10M and 540B. And they are doing so in weeks, not months. This has profound implications for us.

It would seem that Google’s strength—a monolith with access to vast amounts of data and resources—has become it’s weakness:

Part of what makes LoRA so effective is that it’s stackable. This means that as new and better datasets and tasks become available, the model can be cheaply kept up to date, without ever having to pay the cost of a full run.

By contrast, training giant models from scratch not only throws away the pretraining, but also any iterative improvements that have been made on top. In the open source world, it doesn’t take long before these improvements dominate, making a full retrain extremely costly.

Google are also hampered by trying to attack a generic problem, rather than specific use-cases:

LoRA updates are very cheap to produce (~$100) for the most popular model sizes. This means that almost anyone with an idea can generate one and distribute it. Training times under a day are the norm. At that pace, it doesn’t take long before the cumulative effect of all of these fine-tunings overcomes starting off at a size disadvantage.

These models are used and created by people who are deeply immersed in their particular subgenre, lending a depth of knowledge and empathy we cannot hope to match.

1. #artificial-intelligence

(via The FLUX Review, Ep. 98)

One crowdfunding campaign has raised enough money to build remote-operated mounted machine guns that can be controlled with a Steam Deck, a portable gaming computer with joysticks and buttons on the sides (similar to the Nintendo Switch). They’ve given these guns to Ukrainian soldiers, who will be able to swivel the guns around and shoot while not being exposed to gunfire themselves.

1. #technology
2. #ukraine

It’s important to develop a strategy to listen to and act on complaints, harness their benefits, and mitigate their destructive potential. When employees believe their manager doesn’t care about, minimizes, or ignores valid concerns, it can increase stress, decrease engagement, and ignite turnover.

“Telling employees to ‘put a lid’ on [their] feelings is both ineffective and destructive; the emotions will just come out later in counterproductive ways.”

Start with interest and curiosity, consider the intention.

Encourage and help facilitate constructive complaints.

1. #leadership
2. #management
3. #organisational-dynamics

(via Kottke)

1. #mandalorian
2. #music
3. #video

(via Benji)

I used to drive one of these in Japan. The dojo had one as the primary means for the live-in students to move heavy goods and run errands beyond the local village. They are the perfect size.

Unlike new vehicles with onboard computers and complicated proprietary parts, Kei trucks are easy to modify and repair. […] “MotoCheez”, a mechanic from Connecticut, says his YouTube channel’s popularity soared after he started featuring his Kei truck.

1. #agriculture
2. #japan
3. #transport

[astronaut John] Glenn, with the help of NASA engineers, quickly modified the [$40 autofocus, mechanical] camera [he bought at a drug store] to make it usable with his bulky astronaut gloves. They flipped the camera upside down, attached a pistol grip with special buttons to control both the shutter and film advance, and even moved the eyepiece to the bottom, which was now the top of the camera because they had flipped it.

1. #nasa
2. #photography

This is just MAD. He gets pretty confident towards the end.

1. #video

It’s great for rental agents, but not so great for vulnerable renters who find themselves on the wrong end of the algorithms that are can determine who does, and doesn’t, win a tenancy in a tough rental market.

[…] some of the technology already in use is programmed to filter through the data and arrive at what the platform determines are the best prospects. 

It’s a process that leaves people who rent at the mercy of automated decision-making over which they have no control.

In many ways, online rental applications circumvent existing tenancy laws, effectively serving as a tenants check, in much the same way as tenant databases or ‘blacklists’.»

1. #data
2. #privacy
3. #technology

My new favourite background music.

1. #music

These scenarios certainly resonated with me as in many ways they speak to reducing cycle time.

All organisations waste a huge amount of time believing that they are making progress on decisions, when in fact they’re just involved in the theatre of decision making. This happens through indirect actions that feel like progress is being made, but in fact contribute nothing to it. Small changes can speed up progress dramatically.

Tangentially related, I often need to emphasise with my Aikido students the importance of reducing intervals between techniques. Reducing a 15 second changeover to 5 seconds could mean getting in another 10 practice runs.

If, like me, you believe in iterating to learn, reducing cycle time is critical.

1. #decision-making
2. #leadership
3. #software-engineering
4. #technology

Counterintuitive research for 7 points, please:

Ruminants have lower feed use efficiency than monogastric livestock, and produce higher reactive nitrogen and methane emissions, but can utilize human-inedible biomass through foraging and straw feedstock. Here we conduct a counterfactual analysis, replacing ruminants with monogastric livestock to quantify the changes in nitrogen loss and greenhouse gas emissions globally from a whole life cycle perspective. Switching 12% of global livestock production from monogastric to ruminant livestock could reduce nitrogen emissions by 2% and greenhouse gas emissions by 5% due to land use change and lower demand for cropland areas for ruminant feed. The output from released cropland could feed up to 525 million people worldwide. More ruminant products, in addition to optimized management, would generate overall benefits valued at US$468 billion through reducing adverse impacts on human and ecosystem health, and mitigating climate impacts.

1. #agriculture
2. #climate
3. #food
4. #research
5. #science

It only takes a few minutes to induce learned helplessness by exploiting our need for social acceptance.

This can happen through deliberate acts, and it can happen inadvertently as a consequence of poor leadership.

1. #leadership
2. #organisational-dynamics
3. #performance
4. #video

Starting from the point of view that this is probably a management failure – rather than something inherent to the person – correctly puts the onus on the manager for letting someone get into a situation like this. Sure, there are situations where there’s not much the manager could have done, but those are rare.

1. #leadership
2. #management
3. #organisational-dynamics
4. #performance

Well delivered feedback drives improvement, whilst poorly delivered feedback decreases performance and can cause real damage, sometimes lasting for years afterwards.

Good feedback is: Well intentioned; Non-trivial; Truthful; Consensual; Actionable; Timely; Specific; Private; Delivered from your perspective, not that of others; A two-way conversation; Focused; About behaviours and performance, not personalities or style; Combined with positive encouragement.

1. #leadership
2. #management
3. #performance
4. #psychological-safety

A quantum computer in the next decade could crack the encryption our society relies on using Shor’s Algorithm.

1. #privacy
2. #technology
3. #video

One of the things that many facilitators and hosts worry about with complex facilitation practices is the outcomes and the quality of the experience. It is the hardest thing to let go of and probably the last piece of “performative facilitation” that deeply experienced facilitators are able to release.

But that desire and drive for a particular emotional outcome can be as damaging to a meeting as a drive toward a particular material outcome.

Don’t tell people what they will experience. Don’t pre-determine their outcomes or their emotional journey on the day. Let go of that control and enable the environment.

the models tended to produce images of people that look white and male, especially when asked to depict people in positions of authority.

the models’ output overwhelmingly reflected stereotypical gender biases. Adding adjectives such as “compassionate,” “emotional,” or “sensitive” to a prompt describing a profession will more often make the AI model generate a woman instead of a man. In contrast, specifying the adjectives “stubborn,” “intellectual,” or “unreasonable” will in most cases lead to images of men.

In almost all of the representations of Native Americans, they were wearing traditional headdresses, which obviously isn’t the case in real life.

image-making AI systems tend to depict white nonbinary people as almost identical to each other but produce more variations in the way they depict nonbinary people of other ethnicities.

1. #artificial-intelligence
2. #diversity-equity-inclusion
3. #technology

If you write more for internal consumption, consider how to write for two readers: the change for your ideal readers, and assurance for your secondary readers. You might have to change your ideal reader in the middle, but that’s practice. Start with the ideal reader, then add what’s needed for the secondary reader.

1. #communication
2. #writing

The Worry Police aren’t curious; they are concerned.

I frequently have front-row seats for their theatre. I am frequently accountable for unpacking their machinations and translating them into action while simultaneously cleaning up their wake of fear and confusion.

Leaders have real power because they understand how to build trust. They build their judgment with the people who understand the problem and have defensible opinions.

Leaders take the time to understand the situation entirely; they communicate clearly and consistently to their audience.

Leaders are intensely curious. This curiosity creates shared understanding and long-term purpose.