geepytee

Thank you for subbing, i'll try to make this worth your time :)

I also enjoy watching projects from zero, and somehow I'm a fan of long, unedited content. Have you watched George Hotz streams? That's what inspired me in part.

Btw posted part 2 just now - https://www.youtube.com/watch?v=ZGFMXu0In7w

This is so encouraging :)

I just posted part 2 just now - https://www.youtube.com/watch?v=ZGFMXu0In7w

At this point i'm starting to get into programming the servos and figuring out leg kinematics. Unfortunately i'm not an expert but I left in all of my mistakes (no video edits) so hopefully others can learn from them. Or do you think I should just edit mistakes and only show the path that worked?

It's a good point on the binge / getting hooked part. Building a 'library of content' definitely helps.

I just posted part 2 today - https://www.youtube.com/watch?v=ZGFMXu0In7w

This is now 5 hours of content total on just the dog alone, plus my channel has some other random hardware hacking.

Curious, for sharing this on HN, do you think I should be doing Show HN posts instead?

Really appreciate the support! Any feedback on what type of content / rants are actually entertaining?

I just posted part 2 just now - https://www.youtube.com/watch?v=ZGFMXu0In7w

Somehow each new video keeps getting longer, curious to hear if you think that's an issue or it's more like 'as long as the quality is there'

To put in context, I'd need to know how many gallons it normally takes to extinguish similar fires. I do agree that's a large volume of water but unclear if it's comparatively a lot?

While GitHub explores o1, Double [0] just made it available to all users in our VS Code extension, same functionality as copilot. Would love to get some feedback from the HN community! :)

[0] - https://double.bot/

DeepSeek just released this week their new DeepSeek-V2.5 model, which is a "combination" of their DeepSeek-V2-0628 and DeepSeek-Coder-V2-0724 models according to their release tweet. [0]

They claim to surpass GPT-4-Turbo, Claude 3 Opus, and the previous DeepSeek-Coder-V2 model in coding, scripting, and math tasks on their official website [1] and it's fully open sourced [2] with a 128k context window.

It still doesn't show on the LMSYS Chatbot Arena coding leaderboard, which is common with new models, but livebench [3] has it ranked 7th for their coding benchmark, which is the highest ranking for any open source model (not counting mistral-large-2407 as fully open sourced since weights not public), beating meta-llama-3.1-405b-instruct-turbo.

Since this model's strength is coding, I've also made it available for free to anyone who wants to try it as a coding copilot in VS Code [4] (Disclaimer: I'm a co-founder at Double and this is my extension).

Hope others find this as exciting as I do, it's great to see open source models continue to improve!

[0] - https://x.com/deepseek_ai/status/1832026579180163260

[1] - https://www.deepseek.com/

[2] - https://huggingface.co/deepseek-ai/DeepSeek-V2.5

[3] - https://livebench.ai/

[4] - https://double.bot/

> It's not that she's unable to understand the advances, she simply doesn't care.

Honestly that feels deep to me. I wonder to what extend all of the recent LLM advancements are lost on most people.

I also wonder if over a long enough time horizon, the "normies" as you call them catch up, or it goes the other way and the "techies" also start to not care (?)

By performance, is there any stat outside of battery life that really matters here? It is a watch after all

Good question. Don't see it as something to juggle along with building a company, but rather a creative outlet that helps keep burnout at bay. Does that make sense?

Appreciate it, will keep it coming!

I built this website [0] to shine some light into one of the most controversial startup questions, should you get the .com domain for your company, or will an alternative domain do?

Obviously inspired by PG's Change Your Name essay [1] that made a strong case for the .com, but the data interestingly (or predictably) shows a strong trend towards .ai in recent batches.

[0] - https://dotcomvsnot.com/

[1] - https://paulgraham.com/name.html

Not entirely sure that there's interest for this type of long form content, but this is HN after all so if there's any chance of anyone being interested it's gotta be here!

I'm building a robot 'dog' (quadruped with a tail and a head) and documenting every second of it on video. This first part is lots of CADing. I'm being brutally honest and not hiding any of my mistakes.

I also have a background in mechanical engineering so I'm not completely clueless but this project is a first for me. My inspiration was watching George Hotz programming videos (hours long) which I really enjoy and thought there should be a hardware equivalent

Hello HN, I'm the co-founder of Double YC W23 [0].

Been surprised to read that a lot of developers still prefer to copy-paste from tools like ChatGPT and Claude AI when using LLMs to code, instead of relying on retrieval algorithms and blackbox UIs.

Often citing that the lack of visibility makes it hard to understand what parts of their codebase are actually being sent to the LLM, and expressing disappointment that these retrieval mechanisms are not as good as their own manual curation (they know their codebase better than anyone else after all).

We understand that engineering the context inside and across context windows is a huge factor in getting an LLM to produce high quality code. That's why we've built our copilot prioritizing visibility over the context window, and giving the user the tools to carefully curate codebase snippets instead of automating the context retrieval (to this day we do not have automatic context retrieval).

If this describes your experience and want to try it out, Double is free to download and has a generous free trial. At this stage we're trying to get as much feedback as we can get and would appreciate any thoughts you have to share :)

[0] https://double.bot/

This is so cool, thank you for building this!

Hey, my company went through YC in Winter 2023. I'll answer the questions based on my experience, your mileage may vary

1. No daily stand ups but we do have multiple events every week during the batch. Some of these include telling your batch mates and the YC partners about your progress and your next steps to keep everyone accountable.

2. At least once a week during the batch, and you can request more 1-on-1 time if necessary

3. If someone has to force you to find PMF, then as a business owner you are doing something terribly wrong. The decision to pivot is ultimately on you, I've only heard stories of YC partners trying to get you to pivot when you are working on something that's very obviously a waste of time (think tarpit idea).

When I went into YC, our product was a sales tool (https://usedouble.com/), we worked on it through the batch even though progress was slow. The partners never really told us to pivot but had great feedback on what different directions to try.

After demo day, we decided to pivot to an AI coding tool (https://double.bot/) and explaining our rationale to them, and they agreed (can expand more on this, and I've also written about it here: https://docs.double.bot/blog/pivot)

4. In Winter 2023, some of the smaller events were remote. The big in-person events could not be attended remotely.

5. You don't have to do anything but then again, I don't know why you'd choose not to be in SF during the entire batch. This is one of those things were the more you put into it, the more you get out of it. Attend every event, make as many connections as possible, that is how you get the most value out of this experience.

6. I recommend checking out Zoom / Google Hangouts. If your particular industry requires face-to-face meetings, I had a friend who was flying a lot during the batch meeting people (manufacturing space).

7. You may differ your demo day and pitch on the next batch, but I find that most companies choose to focus on their strengths and pitch that on demo day. Most companies have really impressive teams, which can be more appealing that even a nice MRR chart, specially early stage.

Hope this helps! Also sounds like you're working on something interesting and have already found traction. If you'd like an extra set of eyes on your YC application, email me: gonzalo[at]double.bot

Does this include the water inside of living creatures?

This is basically Mistral's attempt at custom GPTs?

This model appears to be full of surprises.

The 50% drop in price for inputs and 33% for outputs vs. the previous 4o model is huge.

It also appears to be topping various benchmarks, ZeroEval's Leaderboard on hugging face [0] actually shows that it beats even Claude 3.5 Sonnet on CRUX [1] which is a code reasoning benchmark.

Shameless plug, I'm the co-founder of Double.bot (YC W23). After seeing the leaderboard above we actually added it to our copilot for anyone to try for free [2]. We try to add all new models the same day they are released

[0]https://huggingface.co/spaces/allenai/ZeroEval

[1]https://crux-eval.github.io/

[2]https://double.bot/

We've been using the term "AI coding assistant"

>As many have witnesses, the Copilot Chat has been just terrible. I give up on my hopes for the current wave of the AI evolution.

Hey, I'd love to hear more about why you think Copilot Chat performs poorly.

I personally had a very similar experience and eventually decided to build a solution for it while in YC (shameless plug but I wrote about the Copilot bugs that annoyed me here: https://docs.double.bot/copilot)

Recently started making Youtube videos (this is my 2nd!). Thought I'd go over this very interesting article on the business of building commercial aircraft, and share some interesting facts and anecdotes.

Lots of thoughts around the wave of AI girlfriend products, the lack of hype around AI pets, and what I think is really exciting in the robotics space (plus some historic remarks of previous successful products).

Giving this YouTube thing a shot, I'm usually more of a twitter thread boy :)

We also added Llama 3.1 405B to our VSCode copilot extension for anyone to try coding with it.

Free trial gets you 50 messages, no credit card required - https://double.bot

(disclaimer, I am the co-founder)

I always wondered how do we get to 'free' energy with our current status quo. As soon as energy prices drop, investment dries off and suddenly future supply disappears (see oil and how many new oil rigs are coming online in the US, see uranium and the current yellow cake shortages)

I'll build a drone for you with open source components if you are shopping for one, it will cost more than a DJI but less than these other options.

Out of curiosity, why did you buy a power meter? Does it help with troubleshooting hardware?

Interesting that he built this after getting closely familiar with Neuralink

The era of drawing machines already came and went, and it wasn't 3D printers that killed them.

I like it! Would like to have some sort of map/visualization of what topics i've explored for when diving into parallel sessions. Right now the UI only supports diving into one branch at the time.