GDD
@gdubdup.bsky.social
Software engineering and AI. Currently @EDF.
Views expressed here are personal and do not represent the views of my employers.
Views expressed here are personal and do not represent the views of my employers.
“tracking everyone, all the time.”
Revealed: Microsoft storing Israeli intelligence trove used to attack Palestinians
The tech giant developed a customized version of its cloud platform for Israel’s Unit 8200, which is housing audio files of millions of calls by Palestinians.
www.972mag.com/microsoft-82...
The tech giant developed a customized version of its cloud platform for Israel’s Unit 8200, which is housing audio files of millions of calls by Palestinians.
www.972mag.com/microsoft-82...
Microsoft storing Israeli intelligence trove used to attack Palestinians
Microsoft developed a customized version of its cloud platform for Israel’s Unit 8200, which is housing audio files of millions of calls by Palestinians.
www.972mag.com
August 7, 2025 at 12:26 PM
“tracking everyone, all the time.”
Reposted by GDD
METR previously estimated that the time horizon of AI agents on software tasks is doubling every 7 months.
We have now analyzed 9 other benchmarks for scientific reasoning, math, robotics, computer use, and self-driving; we observe generally similar rates of improvement.
We have now analyzed 9 other benchmarks for scientific reasoning, math, robotics, computer use, and self-driving; we observe generally similar rates of improvement.
July 14, 2025 at 6:22 PM
METR previously estimated that the time horizon of AI agents on software tasks is doubling every 7 months.
We have now analyzed 9 other benchmarks for scientific reasoning, math, robotics, computer use, and self-driving; we observe generally similar rates of improvement.
We have now analyzed 9 other benchmarks for scientific reasoning, math, robotics, computer use, and self-driving; we observe generally similar rates of improvement.
Reposted by GDD
We ran a randomized controlled trial to see how much AI coding tools speed up experienced open-source developers.
The results surprised us: Developers thought they were 20% faster with AI tools, but they were actually 19% slower when they had access to AI than when they didn't.
The results surprised us: Developers thought they were 20% faster with AI tools, but they were actually 19% slower when they had access to AI than when they didn't.
July 10, 2025 at 7:47 PM
We ran a randomized controlled trial to see how much AI coding tools speed up experienced open-source developers.
The results surprised us: Developers thought they were 20% faster with AI tools, but they were actually 19% slower when they had access to AI than when they didn't.
The results surprised us: Developers thought they were 20% faster with AI tools, but they were actually 19% slower when they had access to AI than when they didn't.
Reposted by GDD
If you use "AI agents" (LLMs calling tools in a loop) you need to be aware of the Lethal Trifecta
Any time you combine access to private data, exposure to untrusted content and the ability to externally communicate an attacker can trick the system into stealing your data […]
Any time you combine access to private data, exposure to untrusted content and the ability to externally communicate an attacker can trick the system into stealing your data […]
Original post on fedi.simonwillison.net
fedi.simonwillison.net
June 16, 2025 at 1:21 PM
If you use "AI agents" (LLMs calling tools in a loop) you need to be aware of the Lethal Trifecta
Any time you combine access to private data, exposure to untrusted content and the ability to externally communicate an attacker can trick the system into stealing your data […]
Any time you combine access to private data, exposure to untrusted content and the ability to externally communicate an attacker can trick the system into stealing your data […]
👓 Coding Without a Laptop - Two Weeks with AR Glasses and Linux on Android (holdtherobot.com/blog/2025/05...)
Coding Without a Laptop - Two Weeks with AR Glasses and Linux on Android | Hold The Robot
I recently learned something that blew my mind;
holdtherobot.com
May 25, 2025 at 11:30 AM
👓 Coding Without a Laptop - Two Weeks with AR Glasses and Linux on Android (holdtherobot.com/blog/2025/05...)
"the muscles for organizational innovation inside companies have atrophied. For decades, companies have outsourced this to consultants or enterprise software vendors who develop generalized approaches that address the issues of many companies at once." 🎯
Individuals keep self-reporting huge gains in productivity from AI & controlled experiments in many industries keep finding these boosts are real, yet most firms are not seeing big effects. Why?
Because gaining from AI requires organizational innovation. www.oneusefulthing.org/p/making-ai-...
Because gaining from AI requires organizational innovation. www.oneusefulthing.org/p/making-ai-...
Making AI Work: Leadership, Lab, and Crowd
A formula for AI in companies
www.oneusefulthing.org
May 24, 2025 at 4:26 PM
"the muscles for organizational innovation inside companies have atrophied. For decades, companies have outsourced this to consultants or enterprise software vendors who develop generalized approaches that address the issues of many companies at once." 🎯
Reposted by GDD
Well gosh. Who could have imagined this would happen? futurism.com/klarna-opena...
Company Regrets Replacing All Those Pesky Human Workers With AI, Just Wants Its Humans Back
Years after outsourcing marketing and customer service gigs to AI, the Swedish company Klarna is looking to hire its humans back.
futurism.com
May 17, 2025 at 9:32 PM
Well gosh. Who could have imagined this would happen? futurism.com/klarna-opena...
Reposted by GDD
Our work deconstructing AGI as a concept is also now a reference in www.aisnakeoil.com/p/agi-is-not....
If you're a podcast person, other coverage includes this podcast from @techpolicypress.bsky.social! www.techpolicy.press/should-agi-r...
If you're a podcast person, other coverage includes this podcast from @techpolicypress.bsky.social! www.techpolicy.press/should-agi-r...
May 8, 2025 at 6:14 PM
Our work deconstructing AGI as a concept is also now a reference in www.aisnakeoil.com/p/agi-is-not....
If you're a podcast person, other coverage includes this podcast from @techpolicypress.bsky.social! www.techpolicy.press/should-agi-r...
If you're a podcast person, other coverage includes this podcast from @techpolicypress.bsky.social! www.techpolicy.press/should-agi-r...
Reposted by GDD
In a new essay from our "Artificial Intelligence and Democratic Freedoms" series, @randomwalker.bsky.social & @sayash.bsky.social make the case for thinking of #AI as normal technology, instead of superintelligence. Read here: knightcolumbia.org/content/ai-a...
AI as Normal Technology
knightcolumbia.org
April 15, 2025 at 2:34 PM
In a new essay from our "Artificial Intelligence and Democratic Freedoms" series, @randomwalker.bsky.social & @sayash.bsky.social make the case for thinking of #AI as normal technology, instead of superintelligence. Read here: knightcolumbia.org/content/ai-a...
Reposted by GDD
1. LLM-generated code tries to run code from online software packages. Which is normal but
2. The packages don’t exist. Which would normally cause an error but
3. Nefarious people have made malware under the package names that LLMs make up most often. So
4. Now the LLM code points to malware.
2. The packages don’t exist. Which would normally cause an error but
3. Nefarious people have made malware under the package names that LLMs make up most often. So
4. Now the LLM code points to malware.
LLMs hallucinating nonexistent software packages with plausible names leads to a new malware vulnerability: "slopsquatting."
LLMs can't stop making up software dependencies and sabotaging everything
: Hallucinated package names fuel 'slopsquatting'
www.theregister.com
April 12, 2025 at 11:43 PM
1. LLM-generated code tries to run code from online software packages. Which is normal but
2. The packages don’t exist. Which would normally cause an error but
3. Nefarious people have made malware under the package names that LLMs make up most often. So
4. Now the LLM code points to malware.
2. The packages don’t exist. Which would normally cause an error but
3. Nefarious people have made malware under the package names that LLMs make up most often. So
4. Now the LLM code points to malware.
Reposted by GDD
Not all AI-assisted programming is vibe coding (but vibe coding rocks) https://simonwillison.net/2025/Mar/19/vibe-coding/
Not all AI-assisted programming is vibe coding, but vibe coding rocks
Vibe coding is having a moment. The term was coined by Andrej Karpathy just a few weeks ago (on February 6th) and has since been featured in the New York …
simonwillison.net
March 19, 2025 at 6:02 PM
Not all AI-assisted programming is vibe coding (but vibe coding rocks) https://simonwillison.net/2025/Mar/19/vibe-coding/
Reposted by GDD
People really want to believe that AI is decimating software jobs, because it feels like poetic justice, but fortunately or unfortunately that narrative is nonsense (so far). 🧵
February 14, 2025 at 8:38 PM
People really want to believe that AI is decimating software jobs, because it feels like poetic justice, but fortunately or unfortunately that narrative is nonsense (so far). 🧵
Reposted by GDD
My 8000-word note on agents: huyenchip.com//2025/01/07/...
1. An AI-powered agent's capability is determined by its tools and its planning ability
2. How to select the best tools for your agent
3. How to augment a model’s planning capability
4. Agent’s failure modes
Feedback is much appreciated!
1. An AI-powered agent's capability is determined by its tools and its planning ability
2. How to select the best tools for your agent
3. How to augment a model’s planning capability
4. Agent’s failure modes
Feedback is much appreciated!
Agents
Intelligent agents are considered by many to be the ultimate goal of AI. The classic book by Stuart Russell and Peter Norvig, Artificial Intelligence: A Modern Approach (Prentice Hall, 1995), defines ...
huyenchip.com
January 7, 2025 at 6:28 PM
My 8000-word note on agents: huyenchip.com//2025/01/07/...
1. An AI-powered agent's capability is determined by its tools and its planning ability
2. How to select the best tools for your agent
3. How to augment a model’s planning capability
4. Agent’s failure modes
Feedback is much appreciated!
1. An AI-powered agent's capability is determined by its tools and its planning ability
2. How to select the best tools for your agent
3. How to augment a model’s planning capability
4. Agent’s failure modes
Feedback is much appreciated!
Reposted by GDD
New AI Snake Oil essay: Last month the AI industry's narrative suddenly flipped — model scaling is dead, but "inference scaling" is taking over. This has left people outside AI confused. What changed? Is AI capability progress slowing? We look at the evidence. 🧵 www.aisnakeoil.com/p/is-ai-prog...
December 19, 2024 at 12:16 PM
New AI Snake Oil essay: Last month the AI industry's narrative suddenly flipped — model scaling is dead, but "inference scaling" is taking over. This has left people outside AI confused. What changed? Is AI capability progress slowing? We look at the evidence. 🧵 www.aisnakeoil.com/p/is-ai-prog...