danjl

13 ms·

151.
by danjl 2y ago
You seem to be considering the funding process from a particular founder perspective. There are lots of different types of founders. Some have the finances to support themselves and don't need prep time.
152.
by danjl 2y ago
The biggest benefit of bootstrapping is that it keeps your focus on your customers and building a product with value, rather than "your next round". I'm also not sure where you get the "3-6 months" timeframe for boo
153.
by danjl 2y ago
Berkeley and Stanford have more or less equal number of VC startups. Some years Cal has more, other years Stanford has more. https://newsroom.haas.berkeley.edu/berkeley-named-top-univer...
154.
by danjl 2y ago
You can extend this to all varieties of PM. Product/Project/Program. All should be replaced by people who actually create things. E.g. I prefer to have a UX Designer handle the Product Manager tasks, since they already think about
155.
by danjl 2y ago
I would prefer `somethingSec`, where "something" indicates the usage better than "time". E.g. `delaySec` or `elapsedSec`.
156.
by danjl 2y ago
Does this problem disappear if you specify your own font for your site, like Roboto?
157.
by danjl 2y ago
So, for people who are paranoid about using the services that are designed for these sorts of sharing, like Dropbox, GDrive, GPhotos, and a million others? These theoretical people would suffer the frustrations of setting up and managing th
158.
by danjl 2y ago
The distributed part of Earthstar seems like the selling point? Otherwise, it certainly does not have the range of database capabilities of either PGlite or SQLite, which also work offline, store data locally, and work in the browser. What
159.
by danjl 2y ago
Click bait headline. Nothing was proven about open source as the future. "To gain a foothold in Western markets, DeepSeek had to open-source its models." This is an opinion, not the only solution. DeepSeek could have remained prop
160.
by danjl 2y ago
Starting a company is all about initiative. Taking action after talking to customers and learning about their problems. AI is about responding to prompts. No LLM takes action on its own. It doesn't define or find target customers or in
161.
by danjl 2y ago
No. Cloud providers, like AWS, Google and Azure, make most of their money off of poorly configured enterprise infrastructure -- unused drives, VMs and K8S cluster. They provide controls that allow you to ramp down resources, but it is up to
162.
by danjl 2y ago
As a developer, the more you understand about the non-technical, business side of your company, the better you can prioritize and communicate about what is important. No need to "read the tea leaves" or guess at how to interpret t
163.
by danjl 2y ago
We already fought and lost this battle with 3D APIs for GPUs. What makes you think that winning strategy would play out any other way for tensor processing?
164.
by danjl 2y ago
If you are building a desktop app, use the OS UI library, e.g. SwiftUI or WinUI3. If you are building a web-app, and you need to deploy to the desktop, use something designed for that purpose, like Electron or Tauri. However, you should a
165.
by danjl 2y ago
Exactly. Not having political battlegrounds makes companies better. Whereas, it sounds like you're saying having people to fight in the battleground is the goal.
166.
by danjl 2y ago
Not taking enough time is a management problem. It doesn't matter whether It is the manager or the developer who takes shortcuts. The problem is planning, not coding.
167.
by danjl 2y ago
> Identifying tech debt before it happens Tech debt is a management problem, not a coding problem. A statement like this undermines my confidence in the story being told, because it indicates the lack of experience of the author.
168.
by danjl 2y ago
So you value passion for the company culture and style over, say, effectiveness, ingenuity, and performance? The idea that you would not just promote the best people because they work at home is just absurd and bad for business.
169.
by danjl 2y ago
The concept that you would rather promote a mediocre employee who works in the office over a star who decides to work at home is just absurd, and horrible for business. I cannot imagine it actually being implemented.
170.
by danjl 2y ago
How does being an individual help as a manager? How does it help to have someone who takes responsibility for mistakes compared to blaming those mistakes on an AI? Does taking responsibility for mistakes make you more money somehow? Maybe y
171.
by danjl 2y ago
I'd go further and claim that it is easier to replace a middle manager, heck even a senior manager, using AI than it is to replace the ICs. As AI context windows increase to the point where they can ingest the important company documen
172.
by danjl 2y ago
It is best to genuflect to our future overlords. They may not forget insolence.
173.
by danjl 2y ago
Also - keep your feet on the ground, or, if needed, on top of a box. Do not sit on your foot. Your back will appreciate your efforts.
174.
by danjl 2y ago
Discord servers seem to be growing in popularity. I think this may be one of the last bastions of human generated data, though eventually the AI bots will take over. The problem is it's really hard to find good servers. Oddly it seems
175.
by danjl 2y ago
I think all the mobile GPUs use UMA. I think the tradeoff point is some complicated function of power envelopes and the benefit of more, though slower, memory vs raw performance at any power or $ cost. Though there are several dozen importa
176.
by danjl 2y ago
+1 for "dumb bubble of poo"
177.
by danjl 2y ago
UMA hurts the GPU too much. Widely parallel processing wants to access memory in bigger chunks than a CPU. If you try to mix access and modification, you lose the benefit of widely parallel processing. Other GPU designers have considered an
178.
Ask HN: Deploying Docker Compose to the Desktop
4 points by danjl 2y ago|0 comments
179.
by danjl 2y ago
Exactly. They can easily sell more Ada 6000s, and I'm pretty sure they would do so rather than sell them for much less as 4090s.
180.
by danjl 2y ago
See, every day we get new data to train LLMs. Just think, all of the current LLMs were trained without Popeye, Tintin, Faulkner and Hemmingway. Just imagine how much better they will be with all this new public domain data to use for traini

More ›