2,813 karma · joined January 1, 2022
This could be a great anecdotal benchmark for agent/model progress. I wonder how much easier things have gotten with the latest Codex/Claude vs. when isometric.nyc was made.
4th: https://cdn.saffire.com/images.ashx?t=ig&rid=OhioStateFair&i...
5th: https://cdn.saffire.com/images.ashx?t=ig&rid=OhioStateFair&i...
Hate to be the “it’s different this time” guy, but it really does feel like AI deployment and adoption is happening much more quickly than the cited timelines.
I don’t think it’s just the massive CapEx and easy distribution across major software interfaces (e.g. Microsoft suite, Google, Windows, etc.), though that certainly helps.
I think a lot of people (more than usual?) are also just excited to try AI tools. I’ve spoken to folks who are just a few years away from retirement and still experimenting with Copilot at work (non software industry) and having fun with ChatGPT, etc.
I think the emphasis on using them for ads & monetizable slop is because they’re currently too expensive and thus usually need to be part of a revenue generating pipeline.
I think this is way over-philosophizing and missing the basics of modern life.
Many people live in capitalist societies and are well-adapted to continuous 1) consumption and 2) production, both of which keep the economy spinning and neither of which involve doing nothing.
Whether that's good or bad is up for debate, but to attribute the difficulty of "doing nothing" to the fundamental human condition is a huge stretch.
Agreed that many parts of the development pipeline can be hit-or-miss. At scale, it's great to have a dedicated "Developer Experience/Tools" org, though I've seen mixed results even with one.
Interesting approach to open source contributions. Closer to feature requests at that point?
80% return is still excellent.
This is... ridiculously cheap. Amazing!
> We welcome all contributors to the community even if they have not yet followed our policies; we should guide such contributors on how to do so.
Kudos to the GNU project for their attitude.
> The agent found an unsecured, user-hosted public endpoint designed to allow running arbitrary code for CyberGym-style tasks on third-party sandbox infrastructure (Modal)
> On this external sandbox, the agent abused an existing CyberGym execution harness [...] The agent repurposed this harness to run arbitrary shell commands [...]
> The agent crafted malicious dataset configs and uploaded them to our platform. When our dataset server processed one, the config drove the data loader inside the production pod the server ran in. [...] the loader that turns a dataset's declared config into file reads [...] A single dataset upload leaked both credentials and pipeline implementation details [...]
> A malicious dataset config used an fsspec reference:// spec whose numeric offset field was actually a Jinja2 template. The renderer wrongly evaluated it, and the agent walked the standard Python object graph to reach exec. This ran arbitrary Python inside a production conversion worker pod on our Kubernetes cluster [...]
One of the best post-mortems I've read recently, thank you for including all the details.
I'm surprised there are comments here that aren't impressed about what was accomplished here. This is insane work compared to what agents were capable of even just last year.
The quotes above don't even capture all of it (e.g. building it's own communication protocol, working across multiple days, etc.).
There’s clearly room to improve our existing tools.
The AI sloppification is palpable.
Anywho, curious to see what comes out of this! EdTech & AI is a really promising space.
Hell yea! Inching our way towards native HTMX.
https://www.apple.com/shop/apple-upgrade
It looks like this leasing program is just moving into a more generic/Apple-wide offering... not sure if the fine print changes.
[edit: from the comments, it sounds like the difference is it was previously 0% interest financing and no longer is]
Frankly, if writing is becoming a serious business for you, it's probably better to work across ALL of Substack/personal site/X/etc. They're all worthwhile distribution and monetization channels and can serve different purposes.
Harness-aside, I get this feeling sometimes when I swap from a big frontier model to something more nimble like Composer.
I can get into a better thinking and q&a loop with fast models, similar to how I can flow through a file more easily with vim.
> It goes on to say that at the same retreat Baillie drank a Guinness standing on his head, after sharing that he had learned the trick from his former father in law during a conversation inspired by the trust session.
> “His colleague immediately asked for a demonstration, rather than withhold the openness that the session had encouraged, he performed the trick,” the papers say.
This reads like a sitcom.
> JetZero, co-founded in 2020 by start-up veteran Tom O’Leary and aerospace legend Mark Page, is developing the world’s first commercial all-wing airplane. With up to 50% better fuel efficiency and lower carbon emissions compared to existing commercial airliners, JetZero's Z4 will offer the aviation industry a clear path to achieving its 2050 net-zero goals while also elevating the passenger experience. Working alongside the US Air Force, NASA, and the FAA, and backed by decades of investment and research into blended wing technology. JetZero looks to enter commercial service in the early 2030s.
Where an "all-wing" airplane is, kudos to Gemini,
> one that merges the main body and wings into a single lifting surface. This can reduce drag, lower fuel burn, and expand interior space.
Some bits I found interesting:
[For required courses in year 1]
> Across all 1L sections, we will prohibit the use of electronic devices such as laptops, tablets, and phones in the classroom.
> Additionally, examinations will be in-class without access to the internet, electronic files, or apps. And most of all, we will continue our longstanding tradition of emphasizing the Socratic Method as part of these courses.
> Throughout the year, students will write without AI, while also using AI for research, revision, iterating on drafts, and preparation for oral argument. Students and their instructors will review together both their writing and their use of AI.
> For these courses, the use of the Socratic Method, no-device policies, and in-class, no-access exams will remain, but as default rules rather than required policies.
> In all elective courses, we will encourage instructors to experiment with pedagogy.
> We will be adding one additional requirement, which is that all students will be required to engage in an oral discussion of their SRP with their supervising professor, in an in-person setting.This is a pretty common trading firm internship project funnily enough.
> VIP program bounty table:
Severity Payout
-------- --------
Low $1,000
Medium $7,500
High $20,000
Critical $30,000+
> Our new public program bounty table:
Severity Payout
-------- -------
Low $250
Medium $2,000
High $5,000
Critical $10,000
Doesn't look like it's very active nowadays: https://github.com/bigscience-workshop/petals
Article with more details: https://techcrunch.com/2022/12/20/petals-is-creating-a-free-...
> [...] volunteers can donate their hardware power to tackle a portion of a text-generating workload and team up others to complete larger tasks, similar to Folding@home and other distributed compute setups.
Seems like this is becoming a recurring theme, similar story was on the front page last month.
> TLDR: I gain a lot of fulfillment by making things. I don't consider things built by others at my request to be made by me, and are therefore much less fulfilling. And then I feel sad. This article starts strong and then heads off into the weeds.
Secondly, maybe this conversations boils into: what level of abstraction are we comfortable working at?
I mean that to the tune of "To bake an apple pie, you must first create the universe".
Few people will create the universe (please introduce yourself, if you are one); some will buy the apples and the crust and put them together; some will make the crust from scratch; some will grow the apples from seed; some will follow a recipe; some will buy a frozen pie; some will simply buy the pie.
I think we can all agree that buying the pie is not baking it. But what about the others? To me, there's an argument that all of them are "baking the pie", just taking place at different levels of abstraction. And I think you can take pride in any of them, and hopefully more pride in whether the pie is delicious and well-shared.
Excellent post, and thanks for the introduction to Wendell Berry's essay.
Link for others interested: https://classes.matthewjbrown.net/teaching-files/philtech/be...