Hacker News new | past | comments | ask | show | jobs | submit | matchagaucho's comments login

My main recollection of 18F and the USDS is that they expanded the number of identity logins on VA.gov from one to three during their tenure.

During one of these transitions, I lost access to my account and had to go through a full re-validation process.

When the VA added ID.me last year as 4th identity provider, and mandated its use by March 4th 2025, it felt like an admission that private-sector providers outperform government-built identity systems.

Internal VA IT resisted years of 18F and USDS-led attempts to refactor the underlying COBOL and MUMPS systems — progress is necessary, but this was not the team to deliver it.


Login.gov is the default idp for the Social Security Administration, supports 200+ federal agencies for identity, and IRS was in the works to onboard Login.gov (to replace ID.me). They handle over 10 million monthly active users and 40 million monthly sign-ins across nearly 50 agencies and states. I would be interested in knowing why it could not serve as the sole idp for VA.gov, as there must be a reason, regardless of validity of that reason.

Appreciate the ground truth, this is enough for someone to dig further in VA from a journalism perspective.


> IRS was in the works to onboard Login.gov (to replace ID.me)

Has anyone agency completed replacing ID.me with Login.gov?

For that matter do any agencies that support both Login.gov and ID.me have a way for users to switch which one they use?

When the ssa.gov first started asking people who used username/password login to add one of Login.gov or ID.me, I picked ID.me because I already had an ID.me account for use with the IRS and did yet have a Login.gov account.

Since then I've gotten a Login.gov account and would prefer to use that with ssa.gov but there doesn't appear to be any way to set that up. There's nothing I can find in account settings to add another login method to an account.


Current guidance from the SSA login flow is to not use Login.gov if you already have an ID.me account. Identity provider transitions are fraught with peril, and I hesitate to provide guidance that potentially locks you out of the SSA website without any recourse or customer service. Therefore, for the time being, my guidance is to keep using your ID.me idp and login until an official update is provided. I will take a note to provide an update if updated guidance becomes available to me through any channel.

Correction: both Login.gov and Id.me will be the IDP options after March 4th changeover.

https://www.va.gov/?next=loginModal


I've noticed sites like Indeed are still showing job listings scraped from our website 5 years ago.

I've not chased down every scraper and submitted a remove request. But I can easily see how 20%+ of job listings are dead-ends.


Welcome back!

And good timing... our brute force cosine similarity scan on Redis-stored vectors needs some help at the core level (ANN, nearest neighbor, k-mean clustering, anything would be an improvement!)


Similarly I've found old-school OCR is needed for more reliability.


I've been using this to OCR some photos I took of books and it's remarkable at it. My first pass was just a loop where I'd OCR, feed the text to the model and ask it to normalize into a schema but I found out just sending the image to the model and asking it to OCR and turn it into the shape of data I wanted was so much more accurate.


Combining google's ocr with llm gives OCR superpowers. Tell the llm the text is from an ocr and ask it to correct it.


That sounds like it could be very dangerous when the LLM gets it wrong...


Depends what you're using it for. If you're relying on OCR, you've already got to accept some amount of error.


>2) pushing the frontier further out will likely get more difficult.

The upside risk is premised on this point. It'll get so cost prohibitive to build frontier models that only 2-3 players will be left standing (to monetize).


There's a growing divide between inference GPUs and training GPUs.

Investors first need to ask what that ratio might look like in 10 years. 10-to-1? 100-to-1? Inference-to-Training

Assuming for each NVDA training GPU sold there are 100 open source / commodity GPUs doing inference, who owns and supplies those data centers and hardware?


It's turtles all the way down. C++ compiles to C which compiles to Assembly...

AI will own the next layer... dynamically generating SQL and code to access existing systems.


Or maybe just delegate us the work. Maybe we are the assembly of AI and we don't know it yet.


To conflate Slack's T&C faux pas with "self host your own LLM" seems like a stretch.

IT Sec and Compliance must read the T&Cs and make better vendor selection choices.


The recommendation is “self host your own collaboration tools”, not “self host your own LLM”.


Same question... ADS-B transponder surely remains on to prevent in flight collisions / maintain vertical sep.

Guessing FAA is just committing to anonymize their public part 135 operators list.


The best companies can churn their metrics quarterly.

Early days of FB were often driven by MAU. Then P95 performance became a priority, leading to higher DAU.


Join us for AI Startup School this June 16-17 in San Francisco!

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: