It’s because AI can debug a programme and people start thinking it can do fitness and health stuff too, but the thing is, there is no “instant-reacting compiler" for health or fitness. Things change over a long time, till then AI would have run out of context or lost the data from its cache, or the user may have got bored and deleted their account.

i actually do it differently

> (1) Let the LLM randomly perturbate the system.

instead of this i ask LLM to what's least likely to improve performance and then measure it.

sometimes big gains come from places you thought are least likely.

For sure! The hypothesis generation gotta be improved. Your take on the "least likely" is interesting. In the beginning of the repo I was having problems with "hypothesis convergence", your idea may be a nice way to introduce the much needed variability

i am opposed to using anything which is not single binary and not using a sqlite db for self hosted things which don't need to scale to millions of users.

Can't wait for Gimp automation, so i can finally start using it!

Any ideas why anthropic is interested in blender funding?

Presumably because they think agents will become the dominant primary users of tools like Blender, and want a seat at the architectural table to help accelerate that & create useful synergies with Anthropic products and models?

The press release calls out the Blender Python API, specifically, which makes sense for agentic use.

> This support will be dedicated towards Blender core development, to maintain and continuously improve foundational features like the Blender Python API

Pretty much spells it out. They have an interest in extending/supporting the ability for Claude/CC to use and interact with Blender. There may be gaps in endpoints that Anthropic needs to enable certain patterns of automated usage.

As literally stated in the second paragraph of the blog post:

> This support will be dedicated towards Blender core development, to maintain and continuously improve foundational features like the Blender Python API, which enables developers and artists alike to extend and improve the software for custom workflows.

https://news.ycombinator.com/item?id=47936552

When in doubt, dogfooding: "make us popular with the Internet crowd, take a look at what popular companies have done. Here's a budget you can use"

Chances are they were expecting the agent to spoon-feed hundreds of influencers.

I think maybe they want to expand 3d creations and modeling for future video gen with avatar type situation or maybe get into 3d game development.

Anthropic CPO was in Figma's board and stayed there a day before Anthropic-Canva Figma-killer came up /s

good PR, probably

3D printer censorship

I just have a Spot instance we use for our builds. It's turned on via serverless, runs it's job with a timeout and exits.

Lately i don't use any managed services and life couldn't be any simpler.

My team has been using https://runs-on.com/ for AWS instance runners, had a few glitches but largely been great for using AWS instances for runners.

I used it, but it prevented my mac from sleeping. After some investigation I found it's local send.

Does it run in the background?

I am operating as an advisor for Indian firms targeting the international market. What I’ve noticed is that many of the companies have moved away from Intercom, Zendesk, or other such tools and built in-house simpler versions.

The same thing applies to Chargbee, Chargify, and other such tools. These ready-made solutions have many features, and they are complex. Most companies only need a subset of those features but the ability to customise them. Making a general-purpose tool is very difficult.

The same thing I’ve noticed for Uptimerobot, PagerDuty, and others.

As a result, I suspect SaaS revenue will drop further and further.

Companies are questioning why we pay $x,xxx per month for a SaaS solution when we can roll it out for some token expense, highly customizable? And it's nothing new, Google has its software managing its internal stuff.

why would i use it over cloudflare?

Fair question. Cloudflare is great if you're protecting web apps behind a reverse proxy. Flowtriq solves a different problem.

If you're running bare metal, hosting infrastructure, or game servers, you can't just point DNS at Cloudflare and call it a day. Your traffic isn't HTTP. You need something that sits on your network, watches raw traffic, and reacts in real time with BGP FlowSpec, RTBH, or cloud scrubbing. Even in this case study, our customer would have been fully exposed with Cloudflare as the attack was across L3/L4 vectors. That's not HTTP traffic you can stick behind Cloudflare's reverse proxy.

Flowtriq runs directly on your servers, detects attacks in under a second, and automatically deploys mitigation at the network level. Full PCAP evidence, instant alerts on Slack/Discord/PagerDuty.

TLDR: Cloudflare is a CDN that proxies your HTTP traffic. Flowtriq is a lightweight agent that lives on your infrastructure, does deep packet inspection on live traffic, classifies attack vectors, and runs automated mitigation runbooks with full analysis and alerting (among many other things). Different tools for different problems.

Instead of burning tokens on SOTA models, why not use a dirt-cheap specialised model for file editing?

Where the SOTA model just makes a cheaper model to make edits, and it does so.

Yeah I also believe that there are plenty of efficiency gains available by using different models for different tasks. Reasoning models such as opus should only be used for the main planning and decision flows, but sub operations (exploring, applying edits etc etc) could be delegated to smaller and cheaper models. You also end up with a much smaller context for the main big model