It is Lunar New Year season right now, 2026 is year of the horse, there is celebratory horse imagery everywhere in many Asian countries right now, so this image could be interpreted as East trampling West. I have no way to know the intention of the person at Qwen who wrote this, but you can form your own conclusions from the prompt:
A muscular, robust adult brown horse stands proudly, its forelegs heavily pressing between the shoulder blades and spine of a reclining man. Its hind legs are taut, its neck held high, its mane flying against the wind, its nostrils flared, and its eyes sharp and focused, exuding a primal sense of power. The subdued man is a white male...
Although there is not an officially-supported fully-offline mode, Figma does spend a significant amount of effort preserving changes made offline if the server disconnects and the tab is closed or crashes. Additionally, pushing out a newer version of the file format from N to N+1 requires a delicate dance between client and server to reload the file, upgrade the stashed changes, and merge. These and many other edge cases are handled as invisibly as possible to the user. A really fun edge case is if user A spends an entire weekend offline working furiously populating some page of a document, user B wonders why there's a seemingly empty page in the document and deletes it, user A comes back online and tries to apply changes to a parent node that no longer exists....
Disclosure: I used to work on some of this when I was at Figma.
In my experience, the UX on "Filter Unknown Senders" is incredibly weak. For example, the red "unread" badge will still be incremented by scam messages that are ostensibly in the "filtered" category. My parents quickly worked around "Filter Unknown Senders" by randomly clicking around until they found "All Senders" to read the scam messages again. The only thing that worked long term was to enable iOS parental controls and block all texts and calls from unknown senders. This requires a bit of up-front work to make sure all the legitimate numbers needed are stored as Contacts. Now our tech discussions are "what is the parental control password" and not "I just gave all my personal information and credit cards to a stranger please help" every week.
Having worked on the performance bits of that code base, I always appreciated that the well-performing parts wouldn't have happened without leadership setting explicit performance goals and allocating resources to pay for monitoring and much engineering time to reach said goals.
If you worked on it ... is there some doc describing the general architecture? It looks like the "frontend" is typescript + c++ and the backend is ... perhaps rust, with rust being a sort of "document server". Like: one instance of the backend code manages one savefile, for potentially many clients that then see eachother's changes.
How does webassembly <> backend communication work? I imagine it's somehow websockets. How do you proxy websocket connections to the appropriate backend? How does the C++ <> rust communication work? Protobuf perhaps?
> How does webassembly <> backend communication work? I imagine it's somehow websockets. How do you proxy websocket connections to the appropriate backend?
I don't think there's much public info about this, besides what you can infer from looking at what goes over the wire (zstd-encoded binary messages). But to your question about routing wss connections to the appropriate backend, we've been working on a proxy/DNS server that does just that (by giving each backend its own ephemeral hostname). We've talked to some Figma engineers and at least on a conceptual level they do something similar (although their routing is path-based rather than DNS-based).
I am so excited to see this live! I always wanted something like this for testing games and complex SPA's. UI's are particularly problematic for writing tests -- there's usually much more churn in that area of an application versus the business logic, and no matter how many tests we did try to write, users (and ambitious internal testers that clicked everywhere like crazy) would find a creative way to break things. I think having UI fuzzing in the CI pipeline will go a long way to deploying UI updates with more confidence. Congrats on launching!
I recently played FixFox, an adorable top-down game where you go around fixing robots and machines in a quirky world where some of the sentient robots formed an anti-repair religion, thereby causing the population to slowly die out. No guns or violence, very wholesome and enjoyable story.
I agree with you, but for me, Dall·E 2 feels good because 90% of the time I can keep hitting the generate button and massage the prompt until I get something inspirational, surprisingly, or visually pleasing. Without access to Imagen, it's impossible for me to compare how much of the "realistic feels" of its images is constrained by the taste of the cherry-pickers.
I've started to ask myself if my own creativity is a result of random sampling from the diffusion tapestry of associated memories and experience on that topic.
From my experiments, the LD one doesn't seem to have been trained on as big or as tagged data set - there's a whole bunch of "in the style of X" that the VQGAN knows* about but the LD doesn't. That might have something to do with it.
Cannot speak directly to experience of original developers, but I could see moving target as being an issue.
Would love to see our code open sourced someday and be useful to others, especially state exchanges, but that is a long-term goal. Currently we are laser-focused on getting everything ready to ship by next open enrollment.
would be curious to get some answers as-to why this wasn't the path to building the website all along?
I mean, the original site was awarded to CGI Inc without any biding process... and they aren't even a US company. Why did we not have US professionals/companies build this thing the first go-round?
I could have imagined a High-Tech All-Stars sort of thing... each major US tech company sends 1 or 2 representatives to collaborate and work together to build this new long-lasting piece of national infrastructure.
The government procurement process is insanely complex. A full competitive bid process for something of this magnitude would have taken many months at least, and would also make it really hard for consortiums to bid because of the up-front cost. Hence the beltway incumbents (Lockheed, Northrop, IBM, and, yes, CGI though they're much smaller) have a much easier time winning these contracts since knowing how to navigate the contracting process and having existing contract vehicles unfortunately gets you a large part of the way there.
I believe CGI actually got this contract via an IDIQ, which is an "indefinite delivery/indefinite quantity" contract, allowing HHS to add additional "tasks" to the IDIQ contract as long as its within certain bounds set by the contract, and if it stays within their GSA fee schedule. IDIQs started becoming popular in the 90s as a more expedient way to contract rather than the months-long or even years-long process of a standard competitive bidding process. However, it only really solved the procurement woes for incumbents. I couldn't find the original CGI-Federal contract, but here's at least their press release: http://www.cgi.com/en/CGI-selected-build-US-wide-competitive...
A muscular, robust adult brown horse stands proudly, its forelegs heavily pressing between the shoulder blades and spine of a reclining man. Its hind legs are taut, its neck held high, its mane flying against the wind, its nostrils flared, and its eyes sharp and focused, exuding a primal sense of power. The subdued man is a white male...
reply