lumost

6 ms·

91.
by lumost 5mo ago
The main story of agent use cases is in enterprise so far. An enterprise will only pay for a model capable of handling the task and no more. Most enterprise's see no need to hire PhDs as factory line workers. Coding is an interesting c
92.
by lumost 5mo ago
Pretty much, Kimi K2.6 is opus 4.6 quality for coding. If you include discounts due to more efficient input caching it is around 1/10th of opus4.6. https://openrouter.ai/moonshotai/kimi-k2.6 The march of cost effi
93.
by lumost 5mo ago
For equal capability tokens, there has been about a 10x drop in cost every 6 months. We are still chasing the best because the best is moving rapidly, but it’s a simple thought experiment to work out what the cost to serve an 8B model from
94.
by lumost 5mo ago
but then you would need to keep running it, what if you don't want to do that anymore? how do you incentivize someone to keep it going to pay off the debt you just borrowed outside of the llc.
95.
by lumost 5mo ago
It provides liquidity to business owners. As a business owner, if you want cash today because you are done with a business. You could go to a bank and get a loan to pay dividends. This is a bad deal for the bank as you have no incentive to
96.
by lumost 5mo ago
How many of those wtf moments are simply from not “being in the room when it happened?” Most enterprise software is riddled with wtf moments demanded as one compromise or another.
97.
by lumost 5mo ago
The great fear of our current incarnation in AI is that it will create a permanent class divide between those who own the AI and those the AI commands. There may not be a "go to college and get a better job" option in the future i
98.
by lumost 5mo ago
This will probably lead to some balancing act like ye olden days of big data etc. Companies want AI native engineers who will use AI to do their work, but don't want AI quality outputs and don't want to drop 200k per year per empl
99.
by lumost 5mo ago
The economics of tokens are still favorable for reasonably well defined tasks at the top tier. I can build out a detailed design and then spend a few k turning it into reality. Compared to spreading such work across multiple engineers the e
100.
by lumost 5mo ago
I have a family split along classic ideological lines between the northeast and southeast of the US. If you are unfamiliar with conservative's hatred toward immigration, I suggest you travel more.
101.
by lumost 5mo ago
Maxed out the 200 dollar per month plan then incurred the overage. This was over the span of 1 month or so.
102.
by lumost 5mo ago
This, I happily used the opus 4.6 fast mode to the tune of 5k for a project. The delivery of the project justified the 5k, if I only spent 500 but delivered the project 1 month later - I would have been in the dog house.
103.
by lumost 5mo ago
In my opinion, this comes from the 70s and 80s where there was very real concern that Japan was going to surpass the US economically. Many companies in the US attempted to adopt Japanese methods in manufacturing and other areas, media then
104.
by lumost 5mo ago
At the size of IPO courting investors is inevitable. You need to make sure that there is enough money available.
105.
by lumost 5mo ago
These companies are burning enough cash that they will need to be public. We’re about to have 3 of the worlds’s largest corporations be massively in the red.
106.
by lumost 5mo ago
Napkin math on 5 year depreciation is 5.5 billion per year for 28 billion. However the 28 billion is cash upfront, spacex is probably paying 10-20% interest on the 28 billion for another 2-6 billion per year. So net you are looking at finan
107.
by lumost 5mo ago
Not my area either! But my understanding is that there are more efficient methods of representing static numbers when you can skip the vram lookup. https://taalas.com/ Is an example startup in this area claiming 16k tok&#x2
108.
by lumost 5mo ago
Capex, opex, quality, and volume are tricky things to balance. On balance, pc/mobile are cheaper to operate than equivalent cloud and on prem deployments. It’s not unreasonable to suppose that in 2 years time an opus 5 quality model wi
109.
by lumost 5mo ago
This probably ends in a deflationary spiral. The ai replaces the jobs, the lack of jobs chills demand, the ai becomes cheaper because it exists in a commodity market. The money printer will be used, and maybe it all works out - or we see we
110.
by lumost 5mo ago
I find it interesting that Mythos was announced the same day that GLM overtook opus4.6 in capability. To me this seems like a careful attempt to cool demand for opensource models which are about to take the overall lead.
111.
by lumost 5mo ago
It's more complicated than this, The US has multiple challenge in its own domestic talent pipeline. In a world of finite slots for elite production and elite employment the US must own the outcome of allocating those slots internationa
112.
by lumost 5mo ago
The 9-5 is doing a major part of that comment. Irregular sleep isn’t the end of the world if you can sleep in and recover. Modern parents don’t get a chance to recover.
113.
by lumost 5mo ago
Honestly, the problem is one of BS detection. Lead engineer says something is not workable? Pm overrides saying that Claude code could do it. Problems found months later at launch and now the engineers are on the hook. New junior onboardee
114.
by lumost 5mo ago
If the project fits on a single server (which it probably can in 2026) docker compose is great.
115.
by lumost 5mo ago
Are we really any different? :) you could easily extend the above story to consolidated images/video/3d environments etc.
116.
by lumost 5mo ago
Currently, it is difficult to live update the model’s parameters in response to new information. This difficulty applies at both an infrastructural level and an optimization level. We simply don’t know how to incorporate new information wit
117.
by lumost 6mo ago
This gets a bit tricky. Over very long task contexts (1M tokens) or with prompt compression (10s of millions of tokens) the model can alter its priors based on updated evidence. This form of knowledge based learning is not necessarily robus
118.
by lumost 6mo ago
the latent space of the LLM when it chooses each token is 10s or even hundreds of GB for each word that it chooses. It's not really useful to look at LLMs from the perspective of its prediction head which is a very small part of the mo
119.
by lumost 6mo ago
Having just been on the job market, my experience was that career pivots are much harder now. I initially intended to transition to a neighboring field after an education break - but none of the companies in that field would speak to me. At
120.
by lumost 6mo ago
I spend 400-500 dollars per day during active development at this point. However with more aggressive task breakdowns I can spend ~5k per day. These spend rates are in part due to operating on a larger code base. Operating on a larger code

More ›