fullstack-dev democratizing intelligence 9mcp inc. ezdoesit inc. github.gg @skunkworks_ai @alignment_lab
Joined June 2008
- Tweets 29,143
- Following 8,141
- Followers 17,882
- Likes 96,449
3,506 Photos and videos
Pinned Tweet
17 Mar 2023
I'd💞to thank @DanielSnider from @VectorInst for helping draft this open architecture for Canadian hospitals to deploy AI in a safe, scalable & auditable manner.
It enables both public & private owned code to interoperably solve the whole problem.
CC3.0 license💫Feedback welcome
35
28
466
139,299
Jun 13
What I speculate happened, is Trump et al used a jailbreaking excuse to try to restrict Claude to US citizens only in order to use it as a bargaining/bullying chip, and (as much as I dont like Dario) I think anthropic did the morally right thing to just cut it out from everyone.
31
7
105
8,587
Jun 13
it just built the most amazing full city drone defense simulator game too :(
2
313
Jun 12
Needed to fire up kimi on opencode to actually check my system for this because claude kept tripping.
The irony of strict AI safety is that it makes models useless at cyberdefense.
🚨 BREAKING: More than 400 Arch Linux User Repository packages have been compromised with infostealer malware and a rootkit.
Attacker posed as a trusted maintainer and "adopted" orphaned packages.
Arch maintainers are purging infected packages now. Audit your AUR installs.
1
1
14
1,366
Jun 12
there's a constant firehosing technique being applied on this particularly via random small instagram influencers
Called it. OpenAI just confirmed something we've been suspecting for a while on ThursdAI.
Are the conversations happening online actually organic?
Absolutely crazy.
1
4
434
Jun 11
bad faith aside, claude fable on my standard Mount Olympus Mons test built the best martian rail launch simulator and closest thing to how i've imaginged the concept years ago.
claude.ai/public/artifacts/6…
2
1
37
3,420
Jun 10
If only our VCs were not dogwater realestate mentality types at their jobs....
1
1
16
1,294
Jun 10
uncensored opensource models are critical infrastructure at this point
Jun 10
NEW: malware developers added nuclear & biological weapons text to to their spyware.
Goal? To trigger LLM safety refusals... so that their spyware wouldn't be analyzed by an AI security scanner.
Cleanest practical example I can think of for why over-indexing on first order safety alignment is risky.
When closed (and open) models ship with aggressive refusals, they will be sprinkled with second-order blindspots that attackers will discover...and exploit.
We are only in the earliest days of attackers leveraging these features, and it wouldn't surprise me if users systems that need to handle complex cybersecurity issues demand that models be less safety-blunted.
In the weeds: @SocketSecurity's post also shows why intention matters in how you design a malware analysis pipeline to avoid prompt manipulation.
H/T to colleagues that shared this with me socket.dev/blog/mini-shai-hu…
3
7
138
3,573
nisten🇨🇦e/acc retweeted
Jun 10
After years of slow development, aMule is back with version 3.0.0, bringing one of the biggest updates in the project’s history.
The release modernizes the classic eD2k/Kad file-sharing client and delivers huge performance improvements.
According to the developers, download speeds can be between 100x and 380x times faster than those in version 2.3.3 on the same hardware.
Some of the biggest changes include:
- Download speeds up to 380× faster than aMule 2.3.3.
- Upload speeds up to 4.8× faster than eMule 0.70b in some tests.
- Disk operations no longer block the main thread, making the client more responsive.
- Upload and download speed limiters have been completely redesigned.
- Better support for very large libraries with more than 100,000 shared files.
- HTTPS support has been restored using modern TLS libraries.
- The project now uses CMake instead of autotools, making development easier.
- Native builds are available for Linux, Windows, and macOS, including ARM64 devices.
- Shared folders are rescanned automatically, and overall UI responsiveness has been improved.
67
152
1,215
67,206
Jun 10
If Claude fable flags a conversation (very easy to happen) they can store that seshion for 2 years and they conveniently forgot to mention if they train on that.
privacy.claude.com/en/articl…
15
33
351
89,860
nisten🇨🇦e/acc retweeted
Jun 10
Albanians for AI 🇦🇱 is my favorite subplot for Eric Adams
Jun 9
The far left’s new war on AI is deeply misguided, and my community and other communities of color will pay the price.
While countries around the world are preparing their young people for the jobs and technologies of the future, some here would rather bury their heads in the sand than equip our children with the skills they need to succeed in a world that is already changing.
We must push back against this fear-driven approach and ensure our communities are not left behind once again. The future belongs to those who prepare for it, not those who pretend it isn’t coming.
1
1
13
1,899
Jun 8
It's funny to have GitHub go down rn while watching this and contemplating moving to svn
3
7
2,060
nisten🇨🇦e/acc retweeted
Jun 7
And I love Canada. And just to be clear I don’t know a single American who doesn’t LOVE Canada except one and he is a complete lunatic. Pray for us.
251
1,441
12,577
113,866