sails

For what it’s worth I tried an experiment where I had a similar harness (where LLMs competed head to head playing snake) and made them aware of their budget and it had very little impact. This was cheaper open weight models

I do this too, the recent MacBook chargers have 2 usb-c and I have 4 of the plug parts that adapt it. Very minimalist but for short trips with single charger it is great

This might if you’re from the US as I’ve collected the ends over time, but I’ve never seen them for sale except the US one: AC Power Adapter Wall Folding Plug Duck Head,US Standard Plug Duck Head Compatible with MaBook Pro Air/Maci Book/Phone/Pod Power Adapter Brick

Is this possibly a replacement for Orbstack?

Great thanks for giving it some thought, I’ll share a video as it’s all quite an early concept but saving me a lot of time on the bench!

Ah no I hadn't seen that, very interesting.

I'm finding a gap just before running those experiments.

The process of updating firmware, doing basic smoke tests on each device and validating it is live, and can function correctly.

Basically the pre-deployment green light that you get on github, but for hardware.

Have you seen or thought about that at all?

Love this, I’m playing around with the cheapo esp32+servos version of this, super fun.

Something I’m working on is a hardware CLI for agents to run experiments, with a “CICD” pipeline that validates everything and means I can delegate more of the experiments to the agents. I wonder if you have any thoughts on this?

The idea is to allow the coding agent to run the full loop of experiments and validations, with vision, audio, button pressing, speaking etc to interact in place of the human

This seemed the closest, of quite a few! https://steamcommunity.com/sharedfiles/filedetails/?id=61053...

Great! I feel well positioned to say this is great :) I’ve been hoping for something like this.

My ideal: slower, more real time, full maps based on actual locations, replay specific naval battles. Multiplayer (maybe it is?), realistic fog of war. I could go on

Edit: a few more. Sail trim is clunky and seems unresponsive or unclear how it maps to physics. Would be better to have trim on one side and steering on the other. Also to turn off tacking with tap. Steering too sensitive. The battles are great. (All on mobile)

What is the best way to give coding agent a full website so that it can see what I see? With animation and design I’m never sure what it gets when I save the website in the browser. Maybe this is suitable?

> but customers kept pulling us toward their data analysis problems

I hear this all the time, I still don’t think it’s a good justification to build a BI tool, but I hope this time it is different.

Product looks cool! I’m hopeful that agents do actually unlock business analytics and we can move on from the BI concept

Edit: a rough explanation of why you get pulled towards data problems is that they are intractable symptoms of upstream process issues. Customer sees a capable startup and co-opts them into trying to solve their tarpit problems. Happens all the time!

primary use is Document processing automation, or Intelligent Document Processing (IDP)

Experienced the same nightmare a few times over with citations.

React/next is limiting, we have rebuilt this for angular and now redid it again using lit for better compatibility. Our old one is very similar to this.

Loading citations for each field across 1000s of pages, colliding citations for all the messy formats, zoom, rotate etc. what a mess!

Great that you took the time to MIT this as it would have saved us many hours, though I think today Fable + Codex makes it pretty quick

I could recreate these in lit as a fork, would be very useful to have the full set

Yes.

I often laugh (cry) at the Kindle Product Manager team who ship nothing but DRM updates.

How about a dictionary modal where the font is the same size as the page text..? Hard to imagine what they do all day, given they do seem to force updates but nothing seems to improve

I think, without much doubt, that AI will be most positively impactful on small business owners.

My experience running a few LTDs is that there is a gap between the accountants and what you need, and running an SME business means you are too busy not to do stupid things and the net effect is lost productivity, less entrepreneurial activity and less growth overall. Dealing with VAT, PAYE, and a million other stupid small things prevents most people from succeeding at running an effective business.

Claude and OpenAI have been surveyed to be most impactful to SMEs, and I think it’s only going to accelerate.

Hopefully this is hugely positive, I see risks, but I don’t see real societal downsides if people get AI to make their basic business operations better, cheaper and most importantly simpler and easier.

I think this is right ( a big gap ) but I don’t think data companies even now what the right shape is for AI.

It’s definitely not old school ETL + dbt + BI tool, it might be something like this, but it’s very early

You can feel what is possible using Gemini speech to speech model, it can do tool calls and is very fast. It lacks somewhat in thinking capability but you can setup a tool call to a smarter model and it acts as a relay. I’ve been very impressed.

What is the best way for someone without a licence to get this working as quickly as possible? I have used CAD before but would like to have Claude code do it all locally from CLI

Thanks. I mean self hosting a shared version of this on internal infra but designed to be slightly collaborative.

How do you anticipate teams deploying this? I’m wary of GitHub for sensitive business documents, and wonder what an easy secure agent friendly deployment looks like. Cloudflare or GCP are maybe good candidates

> Accent, dialect, and low-resource language adaptation — adapt a base Gemma model to underrepresented voices and languages with your own labeled audio.

Is this for TTS? Have been looking for something to do a local fine tune to get a specific accent

I like it. I feel like this is a possible evolution of the browser.

Going further, AI internet browser could be an entirely new app to break from the legacy.

I feel this with coding agents, so often where it fetches web data and interprets it, html in that loop is only occasionally additive. Feels quite futuristic

This is good! Switching from the various terrible online tools I cobble together. (Descript, Riverside, etc etc)

Request for transcription and transcription editing

I’m doing something similar to simulate llms in b2b lending, it’s slightly slower paced but the core mechanisms are using just-bash to analyse business financials and make profitable loans.

I quite like the idea of llms writing more code up front to execute strategies.

I’m currently developing the game mechanics and ELO. Please share anything relevant if it comes to mind

See also speculative cascades which is a nice read and furthered my understanding of how it all works

https://research.google/blog/speculative-cascades-a-hybrid-a...

Always wondered how auth validation works on these. Could I use your serverless ocr?

It is more robust than when I tried the exact thing with structured outputs API and gpt4 era models, it’s not perfect but surprisingly good

I’ve had some success building “text to dashboard” with this using vercel.

I use bash-tool and Vercel sandbox to generate charts (Echarts) or tables (Tanstack table) from json data, and then json-render to render the charts, tables and markdown into a dashboard.

Looking for an iOS app to test this as I’m generally curious about the capabilities of on devices TTS (yet to find an app, but there are loads for text gen)

It can’t be too far off considering Siri and TTS has been on devices for ages

Any recommendations for an iOS app to test models like this? There are a few good ones for text gen, and it’s a great way to try models

How does it compare to Omarchy? The whole space looks extremely interesting, and on the other hand I need reliable