FlyingAvatar

Not at a startup, but at a small company (<10 developers) we regretted not doing a mono-repo.

We had our own GitLab instance and we were all open source enthusiasts and building micro-ish services, so it seemed natural for us to do multi-repo.

Eventually we realized it was creating a lot of unnecessary overhead, as we often were submitting patches for a single ticket to multiple repos which all required separate reviews.

Pretty cool.

What I think is necessary for this to be viable is to be able to iterate on each page via prompts instead of hoping it all generates well in a big bang.

Looking like you're still pointed at your local development instance:

Cross-Origin Request Blocked: The Same Origin Policy disallows reading the remote resource at http://localhost:3800/iwm-api/0.0.1/site-details/generate-we.... (Reason: CORS request did not succeed). Status code: (null).

Any M-series Mac with RAM larger than the model you want to run on it.

Have you explored the non-hostile possibility first?

Giving a failing CEO a face-saving out (even if all they end up with is a firm handshake) might be easier to accomplish and also be more palatable for potential investors.

I would suspect that in these countries labor is cheap enough that replacing it with robots would not be cost effective.

The only SSDs I have personally experienced failing due to wear are ones that I was using in an array for a database. These drives lasted several (5+) years.

While possible (and I am aware of an issue like this killing some SSDs on M1 Macs early on), it's not something I am overly worried about.

You should be backing up anyway.

So running models on a M1 Mac with 32gb+ works very well as the CPU and GPU RAM are shared so you can run some really significant models with with.

Earlier this year, I also went down the path of looking into building a machine with dual 3090s. Doing it for <$1,000 is fairly challenging once you add case, motherboard, CPU, RAM, etc.

What I ended up doing was getting a used rackmount server that is capable of handling dual GPUs and two nVidia Telsa P40s.

Examples: https://www.ebay.com/itm/284514545745?itmmeta=01HRJZX097EGBP... https://www.ebay.com/itm/145655400112?itmmeta=01HRJZXK512Y3N...

The total here was ~$600 and there was essentially no effort building / assembling the machine, except I needed to order some molex power adapters, which were cheap.

The server is definitely compact, but it can get LOUD when it's running heavy load, so that might be a consideration.

It's probably not the right machine for training models, but it runs inference on GGUF (using ollama) quite well. I have been running Mixtral at zippy token rates and smaller models even faster.

From your post, it's a bit unclear what your background is. Were you a tech worker for a non-tech company trying to get into more interesting tech? That would make a bit more sense with the context you have given.

If this is literally your first tech job, and you expect it would possibly take you another 15 months to find another one, I would think very hard about leaving it.

In either case, I would definitely give it more than a week. I personally would say 90 days minimum. If it's clear then that is not a fit, then I would look for another position in the same city. Perhaps you could get the new place to cover your relocation as a condition of employment since you'd be having to pay it back.

Short answer is No.

I highly recommended watching Andrej Karpathy's Intro to LLMs talk, particularly the section on System 1 vs System 2 thinking. Long story short, what you are describing, using more processing to prepare a better response, is something that is an area of interest, but is not currently part of ChatGPT (or any other LLM that I am aware of).

See: https://youtu.be/zjkBMFhNj_g?t=2100&si=jaImuf3UCn6ReTp4

Not the OP, but I would suspect the Mac Studio will retain its value better than an application specific product.

It can work if the other car took more of the damage than it would have compared to a collision with a "squishier" car. Would be interested to see a Cybertruck vs Cybertruck crash.

Yeah, my M1 Pro's battery has 525 cycles on it and has lost 16% of its capacity.

(Edit: Actually, mine is M1 Max, FWIW.)

I thought this was true for me to too (Firefox / macOS), but the comparisons just took a very long time to load.

I echo the experimentation comment.

Also, often choosing the 'right' technology is less important than designing an architecture that is both simple enough and capable enough.

When doing this in my career, I am always considering can it be more simply (i.e. less complexity is easier to grok and is usually less prone to fail) and can it be done more clearly (i.e. such that the intended use of the architecture is obvious).

To replace the sacrificial anode on my water heater you need to:

1. Dig the plug bolt out from the spray foam insulation with some tiny picks. 2. Get a breaker bar to get enough torque to unseat the nut. 3. Strap the tank to a wall to avoid it moving when using the breaker bar.

The manufacturers don't want you replacing them either. It took over an hour to do something that in theory could take less than 5 minutes.

That said, if you have an non-heat pump electric water heater like mine you're probably better of letting it die and replacing it with a heat pump based model.

Ahh, I think your example would have communicated the benefit of what you were sharing more clearly with longer variable names. For single word variables, though, it feels needlessly verbose.

To answer your earlier question about object definitions, in cases where the objects are small, I do think more concise (single line) notation can be more readable. An example, though not a great one because it's data more than it's code:

  items: [
    { type: "a", quantity: "12", price: 123.45 },
    { type: "b", quantity: "7", price: 456.78 },
    { type: "c", quantity: "3", price: 9.00 },
  ]

If the number of properties exceeds say 3, or the names of them are complex, I would lean toward the longer form.

I appreciate the attempt, but don't the explicit parens make it pretty clear?

I feel like if you just remove the spaces around the parens themselves, it's the most clear: ((a + b) / ((c - d) / e))

Nowhere I have worked would pass a code review for what you have proposed.

Assuming you're fixed on not disconnecting:

I would recommend using ollama (or install llamacpp if you're on Windows) to bring along a modestly capable LLM you can run locally such a CodeLlama or Minstrel. They won't be as capable as ChatGPT, but a lot of knowledge gets compressed into these relatively small models.

You have some possible recourse if someone steals your securities without your authorization.

The reason why it's more appealing for criminals to do it with crypto is that claw-backs are essentially impossible.

Bit rot is definitely a problem with HDDs, especially as their bit densities increased. I've heard that bit rot on SSDs also became a thing as well after they started moving from SLC to MLC.

I'd be interested to see a study on how prevalent either of them are.

A bunch of folks mentioning ARMsid, but there are a number of FPGA-based implementations as well:

FPGASID - https://webstore.kryoflux.com/catalog/product_info.php?produ... reDIP SID - https://github.com/daglem/reDIP-SID

Before you can give a price, you need to talk about their expectations.

What's the SLA? Is it business hours in certain time zones? 24/7/365? Simply being able to provide that support is a cost you will need to factor in.

Will you need to provide SSO capability? Finer grained access control? If you expect to sign up additional enterprise customers, it might be worth eating the cost of this, but if not, you might want to account for it.

I'm not familiar with Cloudflare's violating customer rights? Is it different for free vs paid customers?

Write a paragraph describing the concept you are trying to communicate. Include sentences with each of the parts involved and their connections. Start your diagram by translating this paragraph into a picture. Then you can tweak the diagram as necessary.

Your sample prompts should show what the output would be without having to submit one's email. There is no way for anyone to understand what this does otherwise.

Most of my questioning have been practical questions relating to computer science / programming and language learning (English / Spanish). I have asked a lot of other general knowledge questions, but nothing exceptionally deep.

Can you give more detail as to the hallucinations that you found?

At the moment, this will happen in GPT-4 as well.

I guess I see ChatGPT as a very broadly knowledgeable, yet fallible person, not as a database. It would be a very rare person who could come up with the data you asked for off the top of their heads. Granted, it would be much better if it could tell you "I don't know", rather than hallucinate facts.

As far as it being able to explain base concepts in a conversational manner and helping bridge gaps in my own understanding, it has been very helpful to me.

I'm sure there are certain subjects where this will work better than others, but I would guess especially for most K-12 classwork, ChatGPT even with its current warts would be a net gain for many students in helping to increase their understanding just by talking with it.

What grade level is the homework you are creating?