ianm218

14 ms·

241.
by ianm218 6mo ago
They have been exponentially decreasing so far [1] and the Vera Rubin generations chips will be going live that are 35X more efficient in terms of inference/ megawatt [2]. Even with rising prices 10x is possibly conservative. Maybe if
242.
by ianm218 6mo ago
You’re missing the cost curve of inference. What costs $20-50k now will cost 2-5k in a year or two at which point the math is a no brainer. It makes a lot of sense to build products that almost work now or are almost economical and ride the
243.
by ianm218 6mo ago
I think your narratives are a bit mixed up. We went from science experiment ChatGpt 2022 to zero day finder that is too dangerous to release 4 years later. So whoever has been talking about progressive model updates has surely been correct?
244.
by ianm218 6mo ago
> One would otherwise expect the layoffs to reflect the multiple and much hyped "step change" improvements over the past few years. Instead the chart shows a sudden plateau starting a month ago. Can you clarify the theory here?
245.
by ianm218 6mo ago
I feel like you are applying a US specific/ western lense to this. In East Asian countries they’ve built lots of infrastructure around that presumption already and are committed to it.
246.
by ianm218 7mo ago
Did he preach open source all his life? He seems to have been a long time open source builder who has also previously sold a project.
247.
by ianm218 7mo ago
I don't really understand this comment - most people are not this resourceful within the constraints of their own lives.
248.
by ianm218 7mo ago
I think he is implying that their enterprise contracts are all on prem and airgapped? Seems unlikely to me they do that for all their customers but they likely do for the government ones anyway.
249.
by ianm218 7mo ago
> Why do you care so much? If these are really revolutionary tools that vastly optimize work, why bother forcing people to "try new models and best practices"? If AI makes an employee 10X more productive they get a slight pay r
250.
by ianm218 7mo ago
Total addressable market
251.
by ianm218 7mo ago
Obsidian is super popular and is generally local first and device specific. And even so if your starting a note taking app most of those problems like file corruption and image support are largely solved problems. There is also the benefit
252.
by ianm218 7mo ago
Can you expand on this? You definitely don’t need 6 months for a note taking app to be useable it is more you need to compete with the state of the art right
253.
by ianm218 7mo ago
> If we end up in a place where the craft truly is dead, then congratulations, your value probably just dropped to zero I think the craft is going to die and am not thrilled about it. I dont feel like there is a contradiction there
254.
by ianm218 8mo ago
> Yeah this is just straight up nonsense. I was just sharing my experience I'm not sure what you mean. Just n=1 data point. From first principles I 100% agree and yes I was using a CLI tool I made with typer that has super clear --h
255.
by ianm218 8mo ago
I’m not rooting for open AI but OpenRouter is a very self selecting group. Most API users of Anthropocic or OpenAi would just go through the normal API
256.
by ianm218 8mo ago
Not doubting you just sharing my experience - was able to get dramatically better experience for multi step workflows that involve feedback from SQL compilers with MCP. Probably the right harness to get the same performance with the right t
257.
by ianm218 8mo ago
This was my initial understanding but if you want ai agents to do complex multi step workflows I.e. making data pipelines they just do so much better with MCP. After I got the MCP working my case the performance difference was dramatic
258.
by ianm218 9mo ago
> They do not. Do you have anything to back this up? This seems like a shallow dismissal. Claude Code is mostly used to ship Claude Code and Claude Cowork - which are at multi billion ARR. I use Claude Code to ship technically deep dev t
259.
by ianm218 9mo ago
Isn’t there something off about calling predictions about the future, that aren’t possible with current tech, hype? Like people predicted AI agents would be this huge change, they were called hype since earlier models were so unreliable, an
260.
by ianm218 10mo ago
Started working on an application to make it easy to see what parcels in NYC are upzoned with the City of Yes[1] changes that were passed last year. I started off trying to make it a service to help people who are interested in ADU's g
261.
by ianm218 1y ago
I'm in the same boat of some of the other commenters using Claude Code but I have found it atleast a 2X in routine backend API development. Most updates to our existing APIs would be on the order of "add one more partner integrati