deckar01

I am using Copilot in VSCode and it does stream the thinking output to me. At some point it will say something like "Implementing changes..." similar to "Thinking...", but there is no content to expand. ChatGPT and local models always push the code changes in small chunks. Claude used to and at some point changed.

They changed it do all of the changes in a virtual cloud environment, then dump the final result at the end of the response. Before it would stream changes, so if it made a minimal fix, then decided to go off on a tangent you could stop it quickly. Now you have to wait 5+ minutes to get a single line of code out of it just to find out it also refactored everything and burned a stack of tokens. No amount of prompting seems to force it to make incremental changes locally.

You don’t have to use compute to pad the token count.

You should be running your own fork before making pull requests. You don’t have to get other people to use it, you just need to get in the habit of rebasing regularly and cutting releases for yourself. Someday I hope maintainers get better visibility into downstream improvements without the politics of pull requests.

Mitre used to issue CVEs within 24 hours. I am going on 4 months now with no follow up, and no way to tell them GitHub issued a CVE already… I’m pretty sure they were just rubber stamping before. Considering disclosure normally should be coordinated with maintainers, 3rd parties like Mitre don’t seem to have much to offer or much to gain other than being a bottleneck.

You are missing the section on “x402, UCP, and ACP”: monetization. If the end goal is to get a cut of your paid agent traffic, they have a strong incentive to block free access from automated sources.

It looks like the youtube embed is broken. It is supposed to link to a EEV blog video. It is wild how many times someone brings me broken equipment and it turns out EEV blog has already investigated the same issue for the same device.

https://www.youtube.com/watch?v=r-V_Z3bD_PA

They still don’t render the markdown (or LaTeX) it outputs.

Oh, I have to use their app to take the pic. I can’t use my existing photos anymore.

Microsoft used to have an app called Office Lens. It helped color correct and keystone adjust documents scanned with a phone camera. They pushed an update that gutted the app and said this app has been replaced by OneDrive. After installing OneDrive and dodging multiple dark pattern storage upsells, I discovered the OneDrive app doesn’t have any of the document scanning tools. I’m sure someone got a bonus for increasing OneDrive installs though.

It doesn’t render Markdown or LaTeX. The scrolling is unusable during generation. E4B failed to correctly account for convection and conduction when reasoning about the effects of thermal radiation (31b was very good). After 3 questions in a session (with thinking) E4B went off the rails and started emitting nonsense fragment before the stated token limit was hit (unless it isn’t actually checking).

Wait until you see what happens when you type your address into google earth.

https://youtu.be/xicsyakpIL4

It has constraints, it just doesn’t have a declarative constraint solver.

https://build123d.readthedocs.io/en/latest/tutorial_constrai...

Traditionally CAD programs require declaring geometry, then defining constraint relationships between them. That leaves ambiguity. I often create under-constrained sketches in Fusion, then change a dimension, which breaks the geometry in ways that technically respect the constraints.

They designed an imperative constraint system. A lot of constraints are linear, so you can just order the sketches and reference as needed. For circular or bi-directional references you probably have to define some construction geometry first, solve the constrains there, and reference it.

Something I haven’t seen before is their filter syntax for disambiguating solutions. You can express how the desired solution relates to the surrounding geometry. This constrains correctly across large parameter changes and will error when there is no longer a solution rather than switching to an unexpected solution to the constraint.

https://build123d.readthedocs.io/en/latest/tutorial_constrai...

They could get lucky, make a break through in robotics, and vertically integrate power generation into their business model with minimal human labor.

My local community college still had hand drafting as a prerequisite before they would teach me the software as of a few years ago.

You aren’t accounting for managerial politics. A product manager won’t gamble on a large project to lower operating cost, when their bonus is based on customer acquisition metrics.

Industrial manufacturing still uses 2D. They have to annotate tolerances, materials, and processes just as precisely as the geometry.

CAD programs were designed to automate hand drafting processes. Most of the autocad commands made no sense until Drafting 101. It was a full semester of hand drafting, which did feel excessive. Hand writing hundreds of words in an engineering font is just a waste of time.

People were violating the terms of GPL without consequence long before AI. It is very difficult to determine if binaries were compiled from fragments of GPL code.

The places I have found AI most useful in coding is stripping away layers of abstraction. It is difficult to say as a long time open source contributor, but libraries often tried to cater to everyone and became slow, monolithic piles of abstraction. All the parts of an open source project that are copyrightable are abstraction. When you take away all the branching and make a script that performs all the side effects that some library would have produced for a specific set of args, you are left with something that is not novel. It’s quite liberating to stop fighting errors deep in some UVC driver, and just pull raw bytes from a USB device without a mountain of indirection from decades of irrelevant edge case handling.

Mermaid is really bad about cutting off text after spaces, so you have to insert <br>s everywhere. I’m guessing this is getting rendered instead of escaped by your interface. Or just lost in translation at the tokenizer.

Reputation farming -> upvote rings -> black market promotion

It seems to be an indirect attempt to promote their GitHub project. They had Claude make them an “agent” using Bayesian modeling and Thompson sampling and now they are convinced they have heralded a new era of AI.

Even if you tuned two string to ensure that two specific notes on them vibrated at a perfect interval, there are non-multiplicative overtones modulated by resonance with the rest of the instrument. Those intervals are ideals for minimizing dissonance. Practically, the dissonance of 12TET intervals falls below the noise floor of all the other acoustic distortions that give instruments character.

I am imagining first class support for patches in package managers to allow searching for patches and observing their adoption stats.

> If you truly wish to be helpful, please direct your boundless generative energy toward a repository you personally own and maintain.

This is a habit humans could learn from. Publishing a fork is easier than ever. If you aren’t using your own code in production you shouldn’t expect anyone else to.

If anyone at GitHub is out there. Look at the stats for how many different projects on average that a user PRs a day (that they aren’t a maintainer of). My analysis of a recent day using gharchive showed 99% 1, 1% 2, 0.1% 3. There are so few people PRing 5+ repos I was able to review them manually. They are all bots/scripts. Please rate limit unregistered bots.

Resale value. You practically have to pay someone to take an open box chromebook. The secondary market for apple products lasts longer than apple’s software support.

And you are opening issues on projects trying to get them to adopt your product. Seems like spam to me. How much are you willing to spend maintaining this project if those free tokens go away?

It’s your GitHub profile. It looks suspiciously just like the other 10 GitHub users that have been spamming AI generated issues and PRs for the last 2 weeks. They always go quiet eventually. I suspect because they are violating GitHub’s ToS, but maybe they just run out of free tokens.

Dear god, it’s bots all the way down.

Those users all work for companies that sell AI tools. And the first one literally says they let AI fix merge conflicts. The second one is in a thread advocating for 0 code review (which this can’t guarantee) (and also ew). The third is also saying to just have another bot handle merging.