deepGem

I think micro management is often misunderstood. The usual run off the mill micromanagement, that almost every terrible manager follows is what I call micro verification. Double guessing every step you take and adding a verification layer on top of it. In the long run, it erodes the confidence and risk taking abilities of employees and overall completely detrimental to a company.

Real micromanagement is the ability to dive in and solve unsolvable bugs or issues, rally the team until the minutest detail of an issue has been hashed out or remove any micro obstacle. Once a manager shows skin in the game and proves themselves, they often win the minds and hearts of all engineers, who then feel inspired to get into micro details themselves.

Now like a sleight of hand, you have scaled accountability and integrity in your team.

So micromanagement done in the right situation and in the spirit of moving things forward is highly beneficial. In most cases, however micromanagement devolves into this micro verification tool that every mediocre manager uses to justify their existence.

This is cool. I am using body pose angles to determine the correct pose for a pushup/squat/whatever. It will be really nice not to use the camera for this and somehow determine the correct pose only using the CoreMotion data. I will dig into this.

A combination of watch + airpods should probably solve the overall body pose and positioning.

However, people would still want videos for reference and to get visual feedback.Perhaps not. Just an audio feedback supplanted by occasional video should help.

This explodes for search terms dealing with questions related to bugs or issues or how to dos. Almost all top results are YT videos, each of which will follow the same pattern. First 10 secs garbage followed by request for subscribe and/or sponsorship content then followed by what you want.

I have applied to YC with multiple ideas in the past and have been rejected without interview. In my current iteration what I have going against me - solo founder, late 40s, building in a crowded consumer space.

I have no idea what YC looks for and it doesn't matter. I will still apply last minute. Almost all my past applications helped me refine and rethink about my ideas.

It is highly discouraging to hear a reject but that's the life of a founder.

Impressive demo.

I noticed a correction that was done retroactively in the demo

'make that H100 GPUs'

and noticed that there was only one instance of the token GPU. Hence the correction was seamless. Had there been a couple more instances I guess all GPU tokens would have been replaced by H100 GPUs. I guess you could say make that NVIDIA H100 GPUs that would be more accurate but if there were multiple instances and you needed the change only in one instance, not sure how that'd fly. I am nitpicking but this could be a common theme.

The fact this can retroactively change the text and also understand a command is quite brilliant. I don't see any trigger word for a command, so wondering if I needed the command as part of actual text, how would that work ?

There is a counter argument to this - the iPhone shape did not change, the shape evolved. Apple is never known to change shapes or geometry, they evolve from the same geometric construct.

The AVP fitting into standard sized glasses like swim goggles is a possibility but it'll be more like Cyclops visors. I am not sure if it will ever achieve a Rayban form factor.

There is one key issue of keeping lasers aligned for long durations between satellites and even between a satellite to a ground station. There are vibrations in satellites and even a tiny bit of that vibration translates to beam misalignment. Am not an expert though. That could explain the bursts.

So it's hard to sustain the theoretical 100GPS connection for hours let alone days across 2 end points which are in constant motion.

I have no issues saying no to all of these, at the expense of being called a jerk. I am not tipping someone to do a job that they were supposed to do in the first place, especially in a high tax regimes.

You may ask what has taxation got to do with tipping ? Well in a welfare state, the employees are supposed to be paid a moderate wage in accordance with living wages and inflation. I suppose some portion of the 50%+ tax the government rips off from me goes to this. Am not shelling out a tip on top of this.

In low tax regimes, there is an argument to be made for 15-25% tipping but even that doesn't stand true in current climate. The government has intervened and increased minimum wages everywhere. So unless there is some balance, tipping beyond a certain limit is economically illogical and overrated.

Take a standalone feature of a product and make it free

Indirectly this also forces decomposition and building loosely coupled features even at a product level. The speed for delivery and iteration acts as a force function to build pipelines and systems to adapt the new feature into the product should it be a runaway success.

May be the small engineering benefits for a building obsessed dev.

I read somewhere that Google had to increase all it's perks just to increase expenses so as to balance out some small revenue %. Their early growth was that insane. No company has such luxury today. The barriers to entry almost everywhere is levelled. No company will have the profit head start that Google had, at least not in tech.

I don't get it. If civil aviation is beer money why build the 747, the most iconic passenger plane to be every built, only surpassed perhaps by the Concorde. Something has changed at Boeing in the last couple of decades, that has changed the trajectory of Boeing leadership leading to quality issues.

This is all in hindsight. When you are looking forward, with that kind of money, you have to certainly think about risk and diversify. That said, Bill Gates was more or less left with no options but to sell some MS shares. Almost all founder CEOs do this, 1 to de-risk their wealth and 2 to also perhaps to reduce exposure to scrutiny.

Whoever is sitting on 10B is not trying to 100x that money. They are trying to ensure that 10B doesn't go to 1M or worse to zero. The amount of risk you can take dramatically decreases with the increase in the amount of money you manage, which is why Berkshire is sitting on nearly $150B in cash and so is Apple. There are not that many less risky options to deploy 10s of billions let alone 100s.

If you need middle or random access then it’s an array, neither a queue nor a stack :)

I couldn’t understand the animation at all. LRU is much simpler to understand and implement what’s so complicated about a doubly linked list ?

There is so much to learn from this in a product building context. How the musicians align and deliver the notes just by watching the conductor’s expressions, the excitement overall, the positivity and calmness, yet stupendous outcome.

"This may sound implausible to many of you, considering how hard it is to write efficient matrix multiplication/attention kernels, and how much manpower has been put into CuBLAS and FlashAttention. The key here, however, is that transformer decoding has very unusual computational properties. In particular, because of the KV-cache, for BS=1 every single matrix multiplication in a transformer is actually a matrix vector multiplication."

How did they unlock this key ? In retrospect it seems so simple, but without the KV-cache this possibility would not have emerged at all. Hats off !

I cannot upvote this comment enough. I can't stand these blue/white LED headlights. What an abomination in the night. To top this, in India highways are littered with white LED street lights. I really miss cruising at 65-70 mph on a dark highway, with just the reflectors to guide my path, lit by the the pleasant yellow headlights.

The toughness of a pickup truck has been watered down and Musk is trying to build a vehicle that can stand to the definition of tough. That's what I gleam from his talks and publicly available information. For instance - the bulletproof exterior, now I don't know why anyone would want a bullet proof exterior when your windows are not bulletproof. Then I think may be it is to stop some accidental bullets when you are out on a hunting expedition. Just taking a wild ass guess here.

Other than the toughness, I can't find a single reason why he would go with stainless steel.

Sitting here in India and contemplating about the CT, I would buy this vehicle in a heartbeat. It is so menacing, that it would give me peace of mind while driving on Indian roads. The vehicle by itself is an insurance against road rage :).

"And there is how you do something"

Sorry I don't see the 'how' as necessarily appalling.

The less appalling alternative could have been weeks of discussions and the board asking for Sam's resignation to preserve the decorum of the company. How would that have helped the company ? The internal rife would have spread, employees would have gotten restless, leading to reduced productivity and shipping.

Instead, isn't this a better outcome ? There is immense short term pain, but there is no ambiguity and the company has set a clear course of action.

To affirm that the board has caused a split in the company is quite preposterous, unless you have first hand information that such a split has actually happened. As far as public information is concerned 3 researchers have quit so far, and you have this from one of the EMs.

"For those wondering what’ll happen next, the answer is we’ll keep shipping. @sama & @gdb weren’t micro-managers. The comes from the many geniuses here in research product eng & design. There’s clear internal uniformity among these leaders that we’re here for the bigger mission."

This snippet in fact shows the genius of Sam and gdb, how they enabled the teams to run even in their absence. Is it unfortunate that the board fired Sam, from the engineer's and builder's perspective yes, from the long term AGI research perspective, I don't know.

Brockman is with Sam, which makes them a formidable duo. Should they choose to, they will offer stiff competition to OpenAI but they may not even want to compete.

For most of us technologists, figuring out user need is way way harder than doing research or building something. That's the reality. So we tend to take comfort in building technology and figuring out the user need later on.

Even discovering a real problem, a problem that warrants the expense on technology, is unfortunately out of the comfort zone for a lot of technologists. We often assume that the problem is real, or worse hope that the problem be real and jump into solving it right away because that's our comfort zone, building stuff.

There ain't nothing wrong with this attitude or process. In most cases, where technologists have solved real problems, the real problem has been a serendipitous unraveling during the build-iterate-shut process. So the best shot at figuring out a real problem for technologists is not spending time in problem discovery, but in a better ship-iterate-shut cycle. A cycle, in which in one can look at current usage and postulate the future and take rapid decisions on what to build, or what not to build.

Having read the biographies of many tech leaders, I have come to the understanding that the key skill that has set them apart is that exponential growth in their ability to hone the intuition about future demand, in a very short span of time, starting from a barebones MVP.

My read is that the sweet spot price for EV mass adoption is $15k with about just enough range to manage. Like a 200 mile range with extremely fast charging. A smaller battery that can charge in like 5-7 mins. The car can be a bit smaller, 0-60 in 6 or 7 seconds should be good enough.

What this implies is that there needs a greater density of charging stations - which Tesla is anyway working on this. So in a Tesla super charger, if I can charge to 80% in 5 mins and get 150-160 mile range am good. Even if I don’t have a garage at home, this should work with appropriate alerting, routing etc.

An electric car for the non-enthusiasts, for those who just need a car to get stuff done.

What got me intrigued was how did they send the update and what was the size of the update. Sadly no mention of those aspects in the article. Is there any way to find out ?

I was feeling so low today that I just switched to HN to tune things off of my mind, saw this headline and right off the bat, am in a better mood.

What an incredible feat of engineering !

I wonder if there is a big enough market for a middle ground device. A good SLR like camera which has a great sensor and a lens but a good control compromise on the software. Not fully managed PaaS but the VM equivalent, if you consider Nikon D7100 as the bare metal Linux box of the 90s.

At the expense of latency ? when in fact latency is the most important aspect for any search. Any idea on how fast the searches from the client are ?

retrieve the embedding index and to run an indexed search to identify the data to be retrieved. Please bear with the layman like questioning -

So if the data is {obj: "obj1, "data": {"name": "atlas", "embedding": "1123124234" } What is an embedding index ? Is it something like {"1123124234": "obj1"} ?

From what I understand the query will be "geography" whose embedding will be "12311111" and now you have to run a KNN for a match which will return {"name": "atlas", "embedding": "1123124234"}

Not sure where the embedding index comes into play here.

How is indexing a vector different from indexing a varchar or an integer ? If you convert a vector into a byteaarray it should be no different from a bytearray of varchar but for the bytearray contents.

Now if you want to do similarity search you have to measure the distance between 2 or more vectors and that's independent of the indexing. No ?

So any database with sufficient memory should be able to accomplish this as evidenced by the vector similarity search feature of Redis. ( I don't know how Redis folks have implemented vector similarity but they do support KNN search )

Agreed not shelling out 3K for a camera, it's not the Vision Pro. My sweet spot price for a camera like this is $500. A fabulous 35-200mm zoom, a great sensor and all manual. The only other aspect I wonder about is in camera image stabilization. How can that be achieved without sacrificing battery life.

I was an ardent manual film photography fan. Heck, I even bought a medium format camera. The hassle of developing film was too much even in the US. What won over digital is not just the freedom to click a million pictures, but the freedom to run your own darkroom on your PC/Mac.

What I really need now is a compromise. I want the freedom of my own digital darkroom but I need a fully manual and digital camera with an incredible 4-6 month battery life. So cut everything out of the camera that sucks battery - the LCD screen, the motors, just have the sharpest lens and the best sensor, bring back the optical viewfinder. In fact, I wouldn't even mind having film as a backup, in case the battery runs out.

Sure this will add a bit of weight and all that, but this will appeal to those photographers who love the process, and believe that the end result takes care of itself. I can't help but draw parallels to programming but I'll refrain from doing so.

Being at the right place at the right time, perhaps takes a week of backcountry hiking, so be it. I need a camera that doesn't run out of juice, and has zero lag, for that once in a lifetime shot.

On similar lines, we have auto focus and manual focus, 2 extremes why not a middle ground, assisted focus ?. Nikon AI and AI-P lenses had this ideology, but surely that didn't survive.

Rain water harvesting and catchment - while in theory this is a solution, it has not been implemented in any major city at a large scale. For instance in Bangalore, around 200,000 households have implemented rain water harvesting. Why is this ?

Implementation problems - The process and procedure is not as straightforward as 'dig a hole, put some gravel and you are done'. In fact, if you don't follow the correct protocol, you will screw up the ground water table.

Lack of standards - I am yet to come across a protocol and standard for implementing RWH. Every RWH expert has their own protocol. In fact, someone as famous as Aamir Khan touted a revolutionary process of RWH, which proved to be disastrous for the water table.

The consumer need or want - This is the most impactful reason in my opinion. The water crisis is not in your face, it is not an impending problem. Also, having your own source of water or implementing RWH is not something you can flaunt and show off. Perhaps why people spend so much on the apartment/house and so little on RWH. Had there been a consumer need, private enterprises would have swung to action. We probably would have a Swiggy'ish startup solving RWH here.

Corruption, let's not forget the politician driven agendas for not enabling self sufficiency in water supply. If the water problem goes away, that tool is no longer available for elections. Yes, water crisis is a political tool and is a pretty big one. The farming community almost always is driven by this tool to decide whom they vote.

So, yeah while in theory everyone should implement RWH and the water crisis will be solved, this theory does not hold any water in practice. We need a herculean effort driven either by the govt or a non-profit to bring about the awareness in people. Once the consumer need is established, private enterprises will take care of the rest.

I live in Bangalore, one of the cities which is primed for water depletion. In fact, there are areas where there is no drinking water supply at all and water is ferried by trucks (locally called water tankers). Now these trucks ferry water from God knows where but most likely a fresh water source or a borewell. Most of the neighbourhoods in my city are fed by these borewells some going as deep as 1000ft. Some neighbourhoods don't find water even at those depths and they have to depend on these water trucks.

Now I'm not talking about a few hundred or thousand people I am talking about hundreds of thousands of people who live like this. The prime IT areas are all water depleted already.

There is this notion amongst us Indians that things will improve somehow and we will get by. What they don't realise is that the politicians won't give a damn, when water depletion hits critical levels. In fact these water trucks are run mostly by politicians and their local goons.

The ground reality in a city like Bangalore at the very fundamental liveable parameters is horrible and getting worse day by day. This year rains have been below average so we are staring at a water shortage next year even in areas that get fresh water from the regular water supply.

Think of cessation, also scripturally described as the “non-occurrence of consciousness,” like voluntarily inducing the effects of general anesthesia. Consciousness switches off without a trace, while the basic homeostatic operations of the body — temperature, heartbeat, breathing — remain online.

- Isn't this like the nirvana of nirvanas ? I had heard about some of the monks in the Himalayas able to attain these states but had brushed it off as utter nonsense.

Since now we know this is feasible, can we surmise that there is no need for an external chemical to shut off the mind or to even alter the mental state of a person ? Could you for instance get into an alcohol or a drug induced state without drinking or doing drugs - What a world would that be ..