nibab

im currently building a home. one of the biggest issues is FAR which is very much driven by local laws. are you intending on addressing that at some point ?

looks great btw. congrats

thats a narrow interpretation. people can be violent by causing emotional harm. what do you call that?

cool stuff. my comp sci major feels almost completely redundant in this new vibecoding era and i feel like the only way to stay relevant as a programmer is to learn all these compute primitives and become an LLM systems guy.

the tediousness of keeping documentation up to date and the natural tendency towards small attention spans has always come up as a tax on organizational efficiency: complicated org structures, legibility exercises, communication tollgates etc. there is real value in reducing the friction in the former so that the latter becomes less of a burden.

at the same time, context poisoning is a real cognitive problem for humans too and I can't tell you the number of times I've seen irrelevant details become a drag on execution. my fear is that having too much context will only cause bikeshedding and a revisiting of prior decisions.

frankly, our organizational structures were already pretty good at creating mechanisms for eliciting the right implicit context at large scales. it is possible that we're just going to come up with the same mechanisms from first principles...

fabulous work. ive been following you since danswer. you certainly create a lot of value and have been successful in getting the community to cover the long tail of integrations.

its interesting to see how "lock-in" is the main pitch here. all things considered, i don't think "lock-in" is relevant at all unless the activity performed with the tool is highly strategic to the company.

you could argue that some orgs may not want openai/anthropic to have their sensitive data leave the parameter, but im also here to tell you that even the most privacy sensitive companies in the world probably resolve this by having a proxy in between the users and the LLM APIs from the labs.

so where does this leave you ? cost savings from OSS? maybe, but its hard to imagine that we are in the phase of the adoption cycle where companies have become as acutely aware of costs as you think they are.

my 2c - focus on the integrations and see which one gets most traction. that will be your value capture mechanism long-term.

Looks like Mira/Brad Lightcap were under the impression tha gdb was essentially fired from Stripe. Also, interestingly, Ilya seems not to know who is paying his legal bills?!

It was never about price or performance. Price and performance may be things you care about as a hobbyist, but as a business you have a lot of other considerations.

at ng3n.ai ive been using datalab.to for document processing. currently its mostly for conversion to markdown and some extraction.

ng3n is more of a grid-like workflow solution on top of documents. it's a user-facing application geared towards non-technical users that have processing needs.

if there are all these new problems that became solvable, what exactly are they?

id be interested in replacing datalab with extend, but im not sure what avenues that opens for ng3n. would be very curious to learn!

Isn’t that what agents.md or Claude.md is for?

I think the discussion about normie users vs twitter bubble is fascinating. As times goes on, power users are going to have a worse and worse experience.

Most engineers will scoff at the idea of patching up a legacy project if they see the slightest deviation from “best practices”. They will slam their fists on the table and claim that management keeps “piling on tech debt”. They will argue for a total rewrite and dismiss any concern of said rewrite taking years because “this is what it means to have high standards and best engineering practices”. They wear this as a badge of honor and frame the conversation as a question of morality and purity, in which they of course have the upper hand since they are not motivated by petty business concerns such as profit.

Engineers that refuse to acknowledge constraints, whatever the nature of those constraints may be, are not engineers. At best they are ideologues, at worst they are just incompetent. The most pathetic thing you can do is just continuously deny the laws of physics and reality, because it doesn’t suit you at some ideological level.

Truly elegant solutions are those that account for all constraints in the simplest, most concise way. It is those that do more with less.

despite the title..i read this post as an exhaustion with SV culture. everything starting with the personality types that are elevated (ie some flavor of antisocial personality disorder) and ending with the activities that earn you social capital (ie identifying the next anti-consensus big industry).

i've seen this happen with people that have had much much smaller financial success in the industry..or even ones that haven't had any at all. you are either naturally inclined to identify with the culture or you trick yourselves into it so that you may belong.

<insert paragraph about social desire to be connected and how we construct an image of ourselves through others>

the culture of SV today is an amalgamation of Taylorist ideas, Randian objectivism, Utilitarianism etc etc. there is a lot of social capital to be earned by embodying the values of these currents. DOGE is a quintessential representation of this. it is not surprising at all that author had such a visceral reaction to it.

its important to emphasize that there have been very successful companies that have gone against the current (ie Apple), with an emphasis on craftsmanship, obsession with the process, taste-driven vs data-driven decisions and appreciation for things that are outside of profit maximization.

As a former Amazonian (2017-2020) I was a big believer in the leadership principles partly due to how often they were referenced during my time there and partly due to the business world's obsession with Amazon's secret sauce. The company does a very good job at indoctrinating everyone by using them extensively throughout the hiring/performance management lifecycle. In fact, during the interview loop each interviewer is tasked with evaluating whether the candidate exhibits a specific leadership principle.

As many have pointed out, with time you notice the principles being used in all sorts of ways against you depending on the context. The management class is conditioned to tell you that "principles are deliberately in contention with one another"...which gives everyone the necessary cover to spin a particular guiding principle in whichever way suits them in that moment. Or you could buy the kool-aid and pretend that whoever architected these principles was so linguistically adept that they truly figured out the perfect way of articulating a set of contentious principles that taken together distill the exact cultural nuance that Amazon is all about. Wittgenstein is rolling in his grave.

In their current state, the leadership principles are simply a way of defining the lingo for work conversations and providing some sort of framework for decision making (emphasis on framework)...which still has a bounding effect on how things are done at Amazon, albeit in a very very limited way. I do think most people would have trouble coming up with their own decision-making framework if they had to, never-mind articulating or communicating it to their peers. However, it would be preposterous to claim that the Amazon principles have any significant cultural value, at least in the narrow definition that most people commonly ascribe to "business culture" (ie language is also culture, but in this context its more about unique behavior specific to a company).

What drives the culture more than anything at Amazon is the insane growth that the company has seen in the past couple of decades. People there have felt it very deeply and have the battle scars to prove it. Everyone that has been at the company long enough will point to the often counter-intuitive things that one should do to succeed at sustaining this type of growth. Bear in mind this is a different type of business than the other high growth behemoths of the industry and Amazon has its own peculiar aspects..

As far as the article goes, I think that the most important aspect that the author gets right is that. the decline is probably due to the influx of people in middle management that don't have any idea about what it really took to build Amazon into what it is today.

This video is a fake representation of Chisinau. The author deliberately chose the one derelict area (despite being on a central street, it’s the equivalent of being in the Harlem but yet still on Madison ave) that’s been a point of contention between the local authorities and central government for 20+ years. The hotel is being deliberately left unattended by local authorities as a fuck you to the parliament. Nothing more than a bargaining chip.

This is a terrible misrepresentation of Moldova and the progress it has undergone in the past 20 years. Please look at Moldova’s GDP growth since 2000 compared to the rest of Europe and some BRICS countries https://ourworldindata.org/grapher/national-gdp-constant-usd...

It is as close as you can get to India/Chinas growth without leaving Europe. Chisinau is bustling with activity and full of foreigners chasing entrepreneurial opportunities. The leading Eastern European private equity funds are making hundreds of millions of investments in the country and there appears to be a gold rush in all service oriented areas: IT Consulting, Banking, BPO outsourcing etc.

It is lazy, easy and dishonest to draw a portrait of a derelict country. Diving into the essence of what is actually happening behind the scenes takes a lot more effort. The description of a horse carriage in the center of the capital city is outright false. Chisinau is on par with Kiev, Warsaw, Bucharest, Sofia etc, but bustling with opportunity at every corner.

Like many post Soviet countries, Moldova is incredibly corrupt. Fortunately, it now has a former US federal prosecutor leading anti-corruption efforts. The current president and more than half of the current prime ministers cabinet is made up of Ivy League educated technocrats under 40 yrs of age. It is a very exciting place.

I commend you for plugging AI into a commonly used tool in this space vs creating your own platform for building websites with AI. It's refreshing to see someone working back from a great user experience and meeting the users where they are today. This requires focus in the face of constant integration challenges and lots of design restrictions/limitations. Best of luck going forward!

This is great! I wish my PR review tools allowed me to plug in something like this. Hopefully one day we will go back to the world of customizable/plugin-based software. Most of my web tools are very prescriptive about the user experience and dont let me tailor my tools.

I think the fundamental conflict here is that OpenAI was started as a counter-balance to google AI and all other future resource-rich cos that decide to pursue AI BUT at the same time they needed a socially responsible / ethical vector to piggyback off of to be able to raise money and recruit talent as a non profit.

So, they cant release science that the googles of the world can use to their advantage BUT they kind of have to because that's their whole mission.

The whole thing was sort of dead on arrival and Ilya's email dating to 2016 (!!!!) only amplifies that.

the email evidence confirms that infrastructure is one of the strongest moats in this space. From Elon:

"Not only that, but any algorithmic advances published in a paper somewhere can be almost immediately re-implemented and incorporated. Conversely, algorithmic advances alone are inert without the scale to also make them scary."

This is reinforced by the fact that infrastructure / training details are much more sparse in papers than algorithmic adjustments.

It's all about 1. researchers (capable of navigating the maze and carefully chosing what to adopt from the research space) 2. clean data 3. infrastructure

This can be summarized as a bet on the cost per query going down significantly. Last I checked, it costs OpenAI almost 20x more to run a query vs Google. At that price-point, without serious monetization and even assuming GPU prices go down significantly within the next 5 yrs, it's really more a bet on being to financially withstand the competition.

Even though they offer inference, training is their primary focus.

Inference hasn't really picked up revenue-wise (across the space) comparing to training and it's not a great market to be in. As you mentioned, it's crowded and the barriers to entry are minimal. Anyone with experience in spinning up containers and scaling them can offer this servicer. Paradoxically, it's also the market where the big cloud providers are very well positioned to dominate. Spiky and unpredictable workloads is where their bread and butter is. Their whole economic and infrastructural model is pretty much tailored to this traffic pattern.

Training is a totally different ball game. It is a model that is disruptive to big cloud providers given that it follows very different traffic patterns. Training LLMs involves spinning up 100-1000s of machines for a relatively short period of time and with interconnect that doesn't typically exist in data centers. That is a very unique workload. Additionally you need significantly more specialized ML knowledge in tensor parallelism, optimizations, CUDA etc. That is not as common as scaling a container based workload..

Fun fact: Oracle is surprisingly well positioned in terms of their interconnect fabric. Even Microsoft is partnering with CoreWeave for GPU clusters because they dont have as much capacity interconnected in the right way.

The growth of together.ai makes a lot of sense given that 99% of current AI infrastructure spend is in training (vs inference). As far as navigating the product maze, this is where you want to be in AI right now.

The training revenue stream can sustain the company over the next couple of years as it develops other product lines. They get 2 big bonuses as part of this too: 1. training/finetuning llms for a bunch of companies exposes them to the big problems in this area that customers are willing to spend $ on 2. they're building a big distribution network with AI buyers

This is defensible too. They have a horde of big customers/spenders that are likely incapable of undertaking these efforts on their own and have very high switching costs. Something that you can not say about the big llm providers. Together is sort of flying under the radar serving a great market segment/need while there is a lot of expensive battles being fought everywhere else in the AI space.

This is definitely a company to watch. I wouldn't be surprised to see together.ai becoming a top 3 player in this space.

I think there are some killer products, but the surface area here is not that big. I can definitely see how we can improve on all of the above, but at the rate things are developing right now, you're looking at smaller and specialized models as the underlying infrastructure. Keeping the conversation focused on Anthropic, unless there is promise of some specific functions outside this range that require big powerful models, they're in trouble.

There is a world in which the dust settles and the current "era" of AI doesn't actually result in a significant amount of productivity/value creation and capture thereof. Everything rests on the current assumption that emergent behaviors and some semblance of consciousness can extrapolate infinitely. You have to believe that in order to justify the investments we're currently seeing in some of the big players.

There really isn’t a product in text or other modalities that shows me this is a $1T dollar market in 10 yrs (as the valuations would imply).

To play devils advocate, even if there were some killer products, value capture seems particularly tricky. There just aren’t any business models that could sustain a market mass of this size.

Lots of cool demos and productivity software (mind you with a very narrow definition of what productivity is).

If you were to place all “available money” on a continuum, at one end of the spectrum you’d have the “dumb opportunistic money” and at the other end you have “strategic/partnership capital”. The money is very cheap if it comes from people that want to jump on the bandwagon of fomo. It’s more expensive when it comes from strategic partners that know they have something to offer (and when I say cheaper I mean literally how much $ you pay per preferred share).

I would consider most professional investors in AI somewhere in the middle of this continuum. The big clouds have the rare and coveted GPUs which are the lifeline to AI companies. That gives them way better terms than what a VC firm would get.

Several things to note here:

Amazon as a corporate investor - Of course a lot of this is a futures contract on cloud compute. This indicates how much the leadership here thinks the compute will be problem. Money comes way cheaper outside of the big cloud providers (they also know the importance of compute and pull their leverage). This is not a sure bet. While true AGI is probably sitting behind a huge amount of compute, the “products” that are catching on right now are very much on lower end of the spectrum for required model performance. Small models are cheaper and can run on commodity compute. It’s not entirely clear to me that this is a financially sound bet..

Timing - This is an interesting time to do so. That indicates that the company feels that it’s shown some of its best work and right now is the time to bank and on that (so as to make the leap to the next big breakthrough). Openai did so on the heels of ChatGPT. This is somewhat discouraging, because outside of the context length hackery, Anthropic doesn’t have much to show as a differentiator. At best they’re a me-too startup set on the path to be acqui-hired by Amazon when the VC money subsidizing the compute drains up.

Structure/Size - there was a lot of information about the structure of the openai deal. We’re not so clear on what’s happening here. One of the big questions is valuation. Making a similar promise to openai (ie 50% of profit until 100b) would put the valuation of the company in to 10s of billions. Note that this is a very different proposition than a year ago. In navigating the “product maze” we’ve realized that there aren’t that many killer products. Most enterprises are throwing spend in this direction because the board requires you to have an “ai strategy”. At best, we’re talking about capturing all the VC money that’s going into companies with a new angle on knowledge management/search. As I mentioned above, that’s something that’s getting severely commoditized at the bottom of the market. The prospects here are pretty grim .

have you explored using something more lightweight than react ? for example, wouldn't it be easier to generate vanilla html with tailwindcss ?

Great work! I think the barrier to entry here is going to gradually become smaller and smaller (as models become better and we see more multi-modal models) and the big differentiator here is that you will have lots of open source developers building functionality that is relevant to them (vs keeping this closed-source like Vercel).

I am curious how much of the functionality in this area is "write code that does scaffolding and prompt engineering" vs "finetuning and model-level improvements".

this is a layer on top of sendgrid. common functionality that a lot of users have to implement before any sendgrid api endpoint is invoked.

“Which company is private investment going to fund - the SaaS co. with 40% margins and rapid growth or the manufacturing co. scraping 10% margins and 5% CAGR?”

2 of the most valuable companies started in the last 20 years in the US are SpaceX and Tesla. You can still build a huge amount of value with non-SaaS margins.

I think part of the problem here is how structurally unfit VCs are to fund such companies (ie investment horizon and fund lifecycle is 5-7 yrs). THat’s where scale-up financing by the government can make a huge impact.