esteban 🇺🇦 

140 Followers
400 Following
132 Posts
Old but renewed twitter.com/esteban
All things data. I work at Acryl Data. Before that: SRE/Platform Engineering at Redpanda, HBase at Cloudera. In the before time: Tech advisor at the Mexico City Government (2001-2006)

Married to an amazing baker and father of 2 swimmers, BBQ enthusiast, amateur woodworker, I can fix pretty much everything with the right tools. Still can find my way around in differential geometry, General Relativity, Seismology, Satellites and Nukes.
Telegramhttps://t.me/estebangtz
ProtonMail[email protected]

📢 The GUADEC Call for Papers is officially open!
Submit your paper by March 16th via this link: https://events.gnome.org/event/259/abstracts/#submit-abstract

Join us for an amazing conference in Brescia, Italy, from July 24-29! 🇮🇹

GUADEC 2025

Welcome to GUADEC 2025 GUADEC is the GNOME community’s largest conference, bringing together hundreds of users, contributors, community members, and enthusiastic supporters for a week of talks and workshops. About GNOME GNOME is a free and open-source software environment project supported by a non-profit foundation. Together, the community of contributors and the Foundation create a computing platform and software ecosystem, composed entirely of free software, that is designed to be...

GNOME Events (Indico)
super bowels.

Wired has been killing it lately on covering what may well be the biggest security breach this country has ever seen. They obtained an email sent by an intelligence team within the U.S. Treasury Department that referred to Musk's actions to unilaterally slash and burn government agencies as "the single greatest insider threat risk" they have ever faced.

"There is ongoing litigation, congressional legislation, and widespread protests relating to DOGE’s access to Treasury and the Bureau of the Fiscal Service,” the email reads, per the tech-focused outlet. "If DOGE members have any access to payment systems, we recommend suspending that access immediately and conducting a comprehensive review of all actions they may have taken on these systems."

"There is reporting at other federal agencies indicating that DOGE members have performed unauthorized changes and locked civil servants out of the sensitive systems they gained access to,” the ‘Recommendations’ portion of the email continues. “We further recommend that DOGE members be placed under insider threat monitoring and alerting after their access to payment systems is revoked. Continued access to any payment systems by DOGE members, even “read only,” likely poses the single greatest insider threat risk the Bureau of the Fiscal Service has ever faced.”

https://www.wired.com/story/treasury-bfs-doge-insider-threat

https://archive.ph/q26im

Meanwhile, The Washington Post reports the warning came from a Treasury threat center run by the govt contractor Booz Allen Hamilton. The Post reports that late Friday, Booz Allen said it had “removed” a subcontractor who wrote the warning and would seek to retract or amend it. “The draft report was prepared by a subcontractor to Booz Allen and contained unauthorized personal opinions that are not factual or consistent with our standards,” company spokesperson Jessica Klenk said. Booz Allen won more than $1 billion in multiyear U.S. government contracts last year.

https://www.washingtonpost.com/national-security/2025/02/07/doge-treasury-payments-system-warning/

https://archive.ph/lJ3oi

A US Treasury Threat Intelligence Analysis Designates DOGE Staff as ‘Insider Threat’

An internal email reviewed by WIRED calls DOGE staff's access to federal payments systems “the single biggest insider threat risk the Bureau of the Fiscal Service has ever faced.”

WIRED

A large-scale brute force password attack using almost 2.8 million IP addresses is underway, attempting to guess the credentials for a wide range of networking devices, including those from Palo Alto Networks, Ivanti, and SonicWall.

https://www.bleepingcomputer.com/news/security/massive-brute-force-attack-uses-28-million-ips-to-target-vpn-devices/

Massive brute force attack uses 2.8 million IPs to target VPN devices

A large-scale brute force password attack using almost 2.8 million IP addresses is underway, attempting to guess the credentials for a wide range of networking devices, including those from Palo Alto Networks, Ivanti, and SonicWall.

BleepingComputer
"The greatest threat we've ever faced": US Treasury division classifies #DOGE staff as extreme danger https://www.salon.com/2025/02/07/the-greatest-weve-ever-faced-us-treasury-division-classifies-doge-staff-as-extreme-danger/
"The greatest threat we've ever faced": US Treasury division classifies DOGE staff as extreme danger | Salon.com

Award-winning news and culture, features breaking news, in-depth reporting and criticism on politics, science, food and entertainment.

Testing… testing, 123, testing, testing…
As a third experiment, I asked (in https://chatgpt.com/share/bb0b1cfa-63f6-44bb-805e-8c224f8b9205) the new model to begin the task of formalizing a result in #Lean (specifically, to establish one form of the prime number theorem as a consequence of another) by breaking it up into sublemmas which it would formalize the statement of, but not the proof. Here, the results were promising in that the model understood the task well and performed a sensible initial breakdown of the problem, but was inhibited by the lack of up-to-date information on Lean and its math library in its training, with its code containing several mistakes. However, I could imagine a model of this capability that was specifically finetuned on Lean and Mathlib, and integrated into an IDE, being extremely useful in formalization projects. (3/3)

In https://chatgpt.com/share/94152e76-7511-4943-9d99-1118267f4b2b I gave the new model a challenging complex analysis problem (which I had previously asked GPT4 to assist in writing up a proof of in https://chatgpt.com/share/63c5774a-d58a-47c2-9149-362b05e268b4 ). Here the results were better than previous models, but still slightly disappointing: the new model could work its way to a correct (and well-written) solution *if* provided a lot of hints and prodding, but did not generate the key conceptual ideas on its own, and did make some non-trivial mistakes. The experience seemed roughly on par with trying to advise a mediocre, but not completely incompetent, (static simulation of a) graduate student. However, this was an improvement over previous models, whose capability was closer to an actually incompetent (static simulation of a) graduate student. It may only take one or two further iterations of improved capability (and integration with other tools, such as computer algebra packages and proof assistants) until the level of "(static simulation of a) competent graduate student" is reached, at which point I could see this tool being of significant use in research level tasks. (2/3)

[Parenthetical clarifications added - 9/19/2024]

I have played a little bit with OpenAI's new iteration of #GPT, GPT-o1, which performs an initial reasoning step before running the LLM. It is certainly a more capable tool than previous iterations, though still struggling with the most advanced research mathematical tasks.

Here are some concrete experiments (with a prototype version of the model that I was granted access to). In https://chatgpt.com/share/2ecd7b73-3607-46b3-b855-b29003333b87 I repeated an experiment from https://mathstodon.xyz/@tao/109948249160170335 in which I asked GPT to answer a vaguely worded mathematical query which could be solved by identifying a suitable theorem (Cramer's theorem) from the literature. Previously, GPT was able to mention some relevant concepts but the details were hallucinated nonsense. This time around, Cramer's theorem was identified and a perfectly satisfactory answer was given. (1/3)

"Unfortunately, a recent software update was not successful. Your vehicle cannot be driven.

Please call customer support"