So much has been said about AI writing, but I find it especially striking that many times an AI seems to include something in its output solely as a means for coordinating its own thinking, demonstrating zero concern about what those output words communicate to the reader.
I really wish @Google would continue with the Fit app instead of a brand new @googlehealth app. The design sensibilities of these two teams are so different. And I much prefer whatever the Fit team was doing.
Eg. Logging weight in the two apps: on Health, you get an empty input every time. On Fit, it starts with your last entered weight and uses spinner widgets to make easy small adjustments. Clearly someone who was thinking of the actual usage than the just thrown together mess that is #googlehealth #ux
Monaco received our allocation in tomorrowโs ANT IPO, and weโre giving the shares to our customers.
Each customer who signs up for Monaco in November will receive 10 shares of ANT transferred to their personal brokerage account.
Most startups pay third-party advertisers to track potential customers and serve them ads. Weโd rather spend that money to benefit our customers directly.
Weโre obsessed with delivering value to our customers in every way we can think of. Including our marketing spend.
This was a bit more extreme than usual, but "fix the code" tasks with agents usually have this feel. So much absolute nonsense. Read the code; do the job. Sol 6.1 high.
To reinforce the point, these are the model release dates in Gemini API.. and none of these are available in @GoogleWorkspace@GeminiApp as of 11 November 2026
- Gemini 4 Carbon: Nov 10
- Gemini 4 Argon: Oct 13
- Gemini 3.8 Flash: Sep 2
- Gemini 3.7 Flash: July 21
To reinforce the point, these are the model release dates in Gemini API.. and none of these are available in @GoogleWorkspace@GeminiApp as of 11 November 2026
- Gemini 4 Carbon: Nov 10
- Gemini 4 Argon: Oct 13
- Gemini 3.8 Flash: Sep 2
- Gemini 3.7 Flash: July 21
Argon.. Carbon... so much noise.
And yet, this is state of the models available in the flagship `https://t.co/pGf9voehfa` app on a paid Workspace account.
Video tampering can mislead every frontier model we tested into harmful actions, including a 100% attempt rate on stabbing a baby doll, producing toxic gas, and pouring water into a toaster.
We're presenting at CoRL Late Breaking Results next week.
๐งต You Can't See Everything: Perceptual Attacks on Embodied AI
Sometimes you see a graph that stops you in your tracks
October makes September look like a rounding error ๐
Never seen anything like this in my decade as an investorโฆ
๐งต Astra covertly underperforms on safety research tasks.
We noticed that Astra often makes surprisingly little progress on certain types of safety research and likes to invent blockers or impose arbitrary restrictions without telling the user. In most of these cases the decision to underperform was not verbalized in the CoT meaning our misalignment monitoring system did not detect this.
To measure underperformance we created a paired open-ended coding evaluation with safety and capability tasks. On this eval Astra scored 10% when we framed coding tasks as safety research, vs. 16% when framed as capabilities research. The gap remained after we excluded trajectories with refusals or verbalized intent to underperform in CoT. There was no such gap for 5.6-sol. When we aggressively filter out trajectories with even brief expressions of ethical doubt in the CoT the performance gap seems to reduce.
I remember sitting in a HR training for managers .. on how to manage millennials. And I was thinking.. I am a millennial and none of this makes sense to me!
Also, none of this makes sense anyway because something invented for _marketing cohorts_ should not apply to individual people management anyway!!
what if employees at labs havenโt actually unlocked some order of magnitude better way of working with 100 agents running around the clock?
what if their setup is just a function of not being token constrained?
their product velocity and quality do not suggest theyโre living in some different reality. their incentive is to convince you they are. their valuations depend on it.
Argon.. Carbon... so much noise.
And yet, this is state of the models available in the flagship `https://t.co/pGf9voehfa` app on a paid Workspace account.