trcf23

From what i understood, not so much.

Most of the gains came from fixing a bug + hyperparameters with optuna which is supposed to be already quite automatic (you set the list of all the var with values you want to try and voilà). I guess a simple claude code session would fix that in a few minutes instead of a full day.

To me, I guess the main value of Autoresearch would be to test different kind of architectures. It's sometimes hard to know what to choose and it would probably give a nice overview.

Anyone used it for exploratory modeling?

Totally agree. It's really hard not to fall into thoses tactics/traps because most of what we see is noise.

Frankly google sheet will take you a very long way... Attio and most CRM are built for teams with sales pipeline, not a solo founder trying to find PMF. Most of the features aims at describing complex sales processes and collaboration.

So basically what you want to do is to try different takes (cold emailing, phone, events, influenceurs), get your first 50 customers, try to understand the process (who takes the decision, who will pay, who will use, etc) and based on that, design your CRM to match this process.

My 2 cents (saas solofounder with 6000 users): very important not to overengineer everything. As dev it's our natural tendency but your time and marketing effort is probably the biggest leverage you have.

Good luck :-)

Yet most people seem to agree that AI favours skilled dev more than unskilled ones...

Nice! Do you think it would be easy for someone with no hardware experience to build one?

Probably but with n8n you can keep a trace of execution no?

Well for the OpenAi part, there was another HN thread on it where several people pointed out it was a marketing move more than a technical one.

If Anthropic is able to spend millions for TV commercial to attract laypeople, OpenAi can certainly do the same to gain traction from dev/hacky folks i guess.

One thing i've done so far -not with claws- is to create several n8n workflows like: reading an email, creating a draft + label, connecting to my backend or CRM, etc which allow me to control all that from Claude or Claude Code if needed.

It's been a nice productivity boost but I do accept/review all changes beforehand. I guess the reviewing is what makes it different from openclaws

Has anyone find a useful way to to something with Claws without massive security risk?

As a n8n user, i still don't understand the business value it adds beyond being exciting...

Any resources or blog post to share on that?

The question is: what type of mac mini. If you go for something with 64G + +16 cores, it's probably more than most laptop so you can run much bigger models without impacting your job laptop.

If the idea is to have a few claws instances running non stop and scrapping every bit of the web, emails, etc, it would probably cost quite a lot of money.

But if still feels safer to not have openAI access all my emails directly no?

Very nice, thanks! It’s great to be able to play with viz!

For a deeper tutorial, I highly recommend PyTorch for Deep Learning Professional Certificate on deeplearning.ai — probably one of the best mooc I’ve seen so far

https://www.deeplearning.ai/courses/pytorch-for-deep-learnin...

Wow thanks for sharing! Could you explain how you made the specs? Did you already know pretty much everything you wanted to cover before hand? Was one CC session enough to go through it?

In my experience, trying to make a plan/specs that really match what I want often ends in a struggle with Claude trying to regress to the mean.

Also it’s so easy to write code that I always have tons of ideas I end up implementing that diverge from the original plan…

Ah thanks I didn’t know that…

Never thought to check the debug log.

I would say mostly with the agent panel going into panic in longer conversation or not able to proceed for some reason.

Claude code in terminal also has issues rendering. Never have that with vscode or iterm so I guess it’s zed related.

And sometime for eslint it won’t check ts errors then start adding false errors everywhere and become unusable if I don’t start over.

But the point is to have something easy to read both for humans and LLM, no?

It’s harder to read mermaid in a terminal or a markdown file…

They can’t update it though. In docs it makes sense to use that as a basis and have the Llm update it when needed

Lucky you. It still crashes quite often for me and it drives me nuts that my Claude code history is lost every time…

But love the project and been using it for almost 2 years now though

I didn’t know it was a common thing but it’s definitely something I’ve experienced!

Probably a bit of both but I don’t know any other hardware that is that addictive…

Social network are not necessarily bad, even for teens. The issue here is the effort to make any user into a scrolling machine combined to a medium always in your pocket.

And it’s also mostly targeting children/teenagers. As a parent you can add limitations on cinema, binging series. You can’t on TikTok.

I’m quite glad that there is a form of control preventing a company from a different part of the world that don’t really care about the mental health or wellbeing of my kids to creep into their life like that…

As a parent, it’s not a fair fight and I should not have to delegate that to another private company

Doesn’t it also depend on averaging with free users?

That’s just classical OpenAI trying to make us believe they’re closing on AGI… Like all « so called » research from them and Anthropic about safety alignment and that their tech is so incredibly powerful that guardrails should be put on them.

Do they have an impact on the roadmap? Like special features for gov work or specific integration?

And I believe BlockNote uses tiptap/prosemirror, no? Do they also contribute to those “primitives”? That would be a very nice gift to the OSS community

Or it takes a lot of time effort and intelligence to produce good code and IA is not there yet…

Also most of the studies shown start to be obsolete with AI rapid path of improvements. Opus 4.5 has been a huge game changer for me (combined with CC that I had not used before) since December. Claude code arrived this summer if I’m not mistaken.

So I’m not sure a study from 2024 or impact on code produced during 2024 2025 can be used to judge current ai coding possibilities.