PxldLtd

Yep, I went through the exact same pains. We desperately wanted to move to something else and I kept steering us away from Jenkins as I'd experienced its pains at a previous role. We evaluated tons of different options and begrudgingly settled on Jenkins too.

CI is basically a glorified shell script executor with some env variable/credential injection. We don't use many of the plugins aside from the docker and the agent shell one.

I'm not sure I'm sold on this argument considering the docker image for GH Actions is made by "myoung34" and half the actions are community written, have terrible documentation and feel like abandonware too.

That's the neat part, they get richer whether the war is happening or not. Some get way richer when there's a war on.

I totally agree. There seems to be absolutely zero focus on Glass Steagall or Citizens United so I can't see how this actually happens without political revolt at this point.

Yes, the neo-liberal economy we've ended up with has drifted quite far from well-regulated Capitalism. I'd still argue that we owe a lot of our rights to hard-fought socialist policy though.

I'm of the opinion that everything we've discovered is via combinatorial synthesis. Standing on the shoulders of giants and all that. I'm not sure I've seen any convincing argument that we've discovered anything ex nihilo.

We've jumped ship to self-hosted Jenkins. Woodpecker CI looks cool but Jenkins seemed like a safer bet for us. It's been well worth the effort and it's simplified and sped up our CI massively.

Once we got the email that they were going to charge for self-hosted runners that was the final nail in the coffin for us. They walked it back but we've lost faith entirely in the platform and vision.

Try convincing the CTO that this panoply of smaller players will be around for 5yrs or worth the effort migrating to.

I'm at a much smaller outfit now so we have more freedom but I'd dread to think the arguments I would've had at the 4000+ employee companies I was at before.

I love just blitzing oats, carrots, onions as a base and then throwing in anything else like kidney beans or courgettes. Makes great veggie burgers you can just cook in the oven. Takes no time at all and less effort to cook than a beefburger.

OP here. I posted an earlier version of PipeDream about month ago, I’ve just released v0.3.0 which attempts to improve visual-consistency.

PipeDream is a CLI utility that wraps the stdout of terminal games (like Frotz/Zork or Colossal Cave) and generates visualisations using the Gemini API.

The early version just generated isolated images, which felt disjointed. I built a state-aware navigator and an AI director that attempts to solve this by mapping out physical paths and cardinal directions. When you use the new --img2img flag, PipeDream uses the previous scene as a structural guide. If you walk North, the new room maintains the lighting, style, and architectural geometry of the previous room. It makes exploring feel a bit more cohesive than a slideshow.

I also added aggressive smart-caching. If you backtrack to a previously visited room, it bypasses the LLM entirely and loads the cached state, which keeps API costs down per session.

To try the built-in demo (API key required, free tier works fine): pip install pipedream-fiction export GEMINI_API_KEY=YOUR_GEMINI_API_KEY pipedream-gui --img2img

If you want to try it with a real game, you can install Colossal Cave via uv tool install adventure and run pipedream-gui --img2img adventure.

You can find more examples of usage on the repo: [github.com/cpritch/pipedream]

I recommend running it with the --img2img flag enabled. It costs slightly more per generation, but the visual consistency is entirely worth it. I’d love to hear your thoughts, feedback or suggestions.

It looks like a 2015 wordpress template

OP here. I built a simple CLI wrapper that pipes the stdout of text adventures into real-time image generation.

It uses a state-machine so the visuals stay somewhat consistent as you traverse. It's fully playable now but not validated against many games. It generally costs in the cents/pennies per session thanks to caching.

To try the built-in demo:

`pip install pipedream-fiction`

`export GEMINI_API_KEY=...`

`pipedream-gui`

Let me know what you think

This is a really good point. Competitors in this space have a lot of resources so there's a tightrope to walk if you go the OSS core route. Any of these competitors could leverage your core and provide many more features than you could reasonably implement.

BSL/BUSL seems like a good fit for licensing here. It's technically source available instead of open source but just adds the layer that a competitor can't be built using your core. Otherwise the core is free to modify and fork. AGPL might be an option but I fear it would scare off a lot of companies in the space who have policies against AGPL licensed code but you'd get to keep advertising as OSS.

Sounds like you don't have a good process for handling scope changes. I should know, the place I'm at now it's lacklustre and it makes the job a lot harder.

Usually management backs off if they have a good understanding of the impact a change will make. I can only give a good estimate of impact if I have a solid grip on the current scope of work and deadlines. I've found management to be super reasonable when they actually understand the cost of a feature change.

When there's clear communication and management decides a change is important to the product then great, we have a clear timeline of scope drift and we can review if our team's ever pulled up on delays.

No problem :) I asked a friend who's a bit closer to the space and he agrees, definitely Order Book Heatmaps. The speed you're getting would make this a killer feature.

Competitors typically have to snapshot/aggregate because their graphing libraries are heavily CPU bound. Being able visualise level 2/3 data without downsampling is a big win. Also being able to smoothly roll back through the last 12hrs of tick-level history would be really neat too.

I'd say the bare minimum feature set outside of that is going to be:

- Non linear X axis for gaps/sessions

- Crosshairs that snap to OHLC data

- Logarithmic scales, Candlesticks, Heikin-Ashi, and Volume profiles

- Getting the 'feel' nice so that you can quickly scale and drag (people are real sticklers for the feel of these tools)

- Solid callbacks for events for exchange integration, people hate leaving their charts to place an order eg (onOrderModify etc)

- Provide a nice websocket data ingestion pipeline

- Provide an api so devs can integrate their own indicators, some sort of 'layers' API or something.

Sorry if I can't be of more help as I'm just a hobbyist in this area!

https://caniuse.com/webgpu latest Android Chrome should have WebGPU support. You might need to update.

Have you thought about leaning into some of the fintech space? They'd happily pay for the sorts of features they need to stream financial data (which is usually bazillions of data points) and graph it efficiently.

Off the top of my head, look into Order Book Heatmaps, 3D Volatility Surfaces, Footprint Charts/Volatility deltas. Integrating drawing tools like Fibonacci Retracements, Gann Fans etc. It would make it very attractive to people willing to pay.

I'm sorry but this is factually incorrect and I'm not sure what experts you are referring to here about there being concensus on this topic. I would love know. Geoffrey Hinton, Demis Hassabis, and Yann LeCun all heavily disagree with what you claim.

I think you might be confusing creation ex nihilo with combinatorial synthesis which LLMs excel at. The proposed scenario is a fantastic testcase for exactly this. This doesn't cover verification of course but that's not the question here. The question is wether an already known valid postulate can be synthesized.

I've literally never gotten a promotion without taking on the additional responsibilities first. I wouldn't expect a promotion for just doing time at a company like a prison sentence. If they didn't promote me then I would have immediately moved on.

Almost all the laws are great guidance for Software dev to be honest and idiomatic to a lot of what's banded around today as good practice.

Yeah the claim that tables are hard to style for such a simple design doesn't really hold up in my opinion.

This took two seconds to make https://codesandbox.io/p/sandbox/5ry4rl

I do agree though that subgrid makes the HTML more readable though so I'm all for it.

I think you may be misinterpreting the point. It's not that we never need less regulation, this may be the case. We should never make 'less regulation' the target. The right regulation may be less in some cases.

I thoroughly disagree, I believe world models will be critical in some aspect for text generation too. A predictive world model you can help to validate your token prediction. Take a look at the Code World Model for example.

Here's one I made in Veo3.1 since gemini is the only premium AI I have access to.

Using this image - https://www.whimsicalwidgets.com/wp-content/uploads/2023/07/... and the prompt: "Generate a video demonstrating what will happen when a ball rolls down the top left ramp in this scene."

You'll see it struggles - https://streamable.com/5doxh2 , which is often the case with video gen. You have to describe carefully and orchestrate natural feeling motion and interactions.

You're welcome to try with any other models but I suspect very similar results.

I think a good test of this seems to be to provide an image and get the model to predict what will happen next/if x occurs. They fail spectacularly at Rube-Goldberg machines. I think developing some sort of dedicated prediction model would help massively in extrapolating data. The human subconscious is filled with all sorts of parabolic prediction, gravity, momentum and various other fast-thinking paths that embed these calculations.

That doesn't prevent supply chain attacks. Dev dependencies are still software dependencies and add a certain level of risk.

This is what I think is the next evolution of these models. Our brains are made up of many different types of neurones all interspersed with local regions made up of specific types. From my understanding most approaches to tensors don't integrate these different neuronal models at the node level; it's usually by feeding several disparate models data and combining an end result. Being able to reshape the underlying model and have varying tensor types that can migrate or have a lifetime seems exciting to me.

Anecdotes are not a very useful tool in discussions about generalisations. They provide little evidence aside from saying that it's a category of event that can exist. No one at any point has said citations are gospel. Just that anecdotes aren't adding much to the discussion at hand. If you've got issues with the cited data, be precise instead of casting general aspersions on academia.

Yep, we've got the camera mounted vertically above the screen. We use the dance floors to render an interactive Unity game. The pixel density is pretty good. These panels have an array of IR LEDs embedded in them for detection but the density isn't good enough for small objects we want to track. Great for detecting anything the size of a foot or bigger.

We use IR Floodlights + IR absorbing film + a high speed IR camera. Works great in most lighting conditions and the screen doesn't show up on the camera.