Failing to put sufficient data controls in place would be extremely bad for business. This extends to agents, not just employees.
Code is a small, shrinking fraction of the cost of building a business. Your code isn’t remotely worth getting sued and subsequently losing all their B2B customers over.
If your source code can’t touch any 3p server, self-host agents on open-weight models. If you’re good with public cloud you can run on bedrock or equivalent.
In practice there are good reasons to build these things in-house, but risk of providers stealing your code is pretty low on the list.
This reads to me like a concession by DoW. The latest threat is to cancel Anthropic’s contract, which is what Anthropic was suggesting all along, and much less drastic than alternatives (labeling Anthropic a supply chain risk or invoking DPA).
Yeah. I did 2.5 days in Chongqing in January and didn’t feel like I’d have gained much by staying longer.
Chengdu has more to see and do, even just in the city but especially if you do a day trip to a panda base or something.
All awesome cities. If it's your first time in HK I would do 2 days there, maybe 2/4/2 for an 8 day trip.
and free trade
and really really hate the Constitution
Yes, and its supposed impact on Trump’s reelection in 2024. Slowing relative wage growth in 2025 did not cause people to vote for Trump in 2024.
These are real (inflation-adjusted) growth rates, not wages themselves. Wages for the bottom quartile are now growing more slowly than other groups, but they’re still growing faster than inflation. If they were falling, the values on this graph would be below 0.
Most likely IMO is swarm mode/TeammateTool which is currently behind a feature flag github.com/mikekelly/cl...
(there's a video in the tweet linked from the repo)
I wrote the speech I *wish* Chuck Schumer would give tonight - as an actual opposition leader. Here it is:
"My fellow Americans: At this hour, an unrestrained force of militarized and violent federal officers is carrying out a project of ethnic cleansing in the streets of American cities."
1/14
okay hear me out
Wrote a quick response to the latest anti-YIMBY research paper going around publiccomment.blog/p/sure-it-wo...
I was thinking more of big batch ML jobs like ranking (to the extent non-real-time), labeling, forecasting, etc., where consumers should gracefully handle missing output/fall back to previous run.
Agree your example is still "production" (I would also call it "online") and needs on-call at scale.
Reliability/uptime is one reason you might want a SWE skillset in ML, but it’s not the only reason. Offline ML systems at scale can benefit a lot from performance optimizations and MLOps practices broadly, but will generally still have looser SLAs than online services, often meaning no oncall.
a lot of people don't realize that Christmas carols and traditions are actually folk memories of suppressed pre-Christian events, namely the end of the Third Age and the Fall of Sauron.
incredible idea
dosaygo-studio.github.io/hn-front-pag...
Whether this is a good thing depends on how likely congress is to pass meaningful AI safety regulation.
I am personally less-than-infinitely pessimistic about this, which is more than I can say about most things that require congressional action.
I read this as an acknowledgment that (enforceable) federal preemption of state-level AI safety regulation is only going to happen if it comes with meaningful federal AI safety regulation.
Earlier this year, I would have been surprised by this result.
Congratulations Sean O'Brien and #teamsters!
a “Space Force”, if you will
Note that all of the specifics here apply only to Waymo, not to self-driving cars in general.
I increasingly worry that regulators and the public will paint all platforms with the same brush, and was pleasantly surprised that that didn't seem to happen after the Cruise incident.
19/19
The reason I only say "fairly high" doesn't show up in the data at all - it's ~entirely due to the risk of a bad software update, security incident, etc.
This is unlikely, but it's hard to say exactly how unlikely, and it could be very high-severity if it happened.
18/
FWIW, I have fairly high confidence that Waymo is safer than driving or rideshare in all of its established metros, though not by enough to justify the cost from a pure QALY/$ perspective.
(This is based on my experience of Waymo being ~$5-$15 more than Uber/Lyft depending on distance.)
17/
From a measurement perspective, it's fine to use a metric that allocates >0 fatalities to Waymo for each of these incidents.
From a practical perspective, if these incidents cause you to think Waymo is less safe than you previously believed, I think that's the wrong conclusion.
16/
A nice thing about small datasets is that you can just look at the individual datapoints.
In both of the fatal Waymo-involved accidents to date, the Waymo was rear-ended while at a standstill or near-standstill: once in traffic, once while yielding to a pedestrian before a right turn.
15/
The caveat is that even a perfect driver (which, to be clear, Waymo is not) would be responsible for >0 fatalities in expectation based on a blameless approach.
That doesn't mean it's wrong - just a caveat that needs to be kept in mind.
14/
Now, returning to the 0.47 allocated fatalities
There's a lot to like about the author's methodology, for exactly the reasons he points out. In fact, I'm inclined to agree that blame shouldn't be accounted for in safety metrics.
13/
You could come up with a hierarchical approach to formalize this, but informally: a sufficiently-powered 79% reduction in crashes plus an underpowered 70% reduction in fatalities together provide practically significant evidence of a reduction in fatalities.
12/
But Waymo's report from this January (through 56.7M VMT, doi.org/10.1080/1538...) consistently shows >70% reductions in crashes.
The headline number is 79% fewer any-injury crashes (yes, regardless of fault). This one is highly stat sig because crashes are much more common than fatalities.
11/
