Wall Street systems engineer turned consciousness explorer. Author of Laws of Mystery. Building Noescence — where Structure meets Mystery. https://xitter.cf/t.co/MEkMpjWfhD
Azores, Portugal
Joined September 2016
- Tweets9.2K
- Following1.1K
- Followers3.8K
- Likes3.3K
Pinned Tweet
The world as we knew it has disintegrated. Every tool we reach for — science, strategy, politics, even spirituality — was designed for a reality that is no longer here. So now what?
Aleksandr Bulkin retweeted
Today I'm making a contribution to the AI-agent-focused open source community: git-viewspace, a way for multiple agents to share the same git repository cheaply. (Link in comments)
When several AI agents work on the same codebase, giving each one a separate clone is straightforward—but it duplicates data and leaves you coordinating branches, commits, and integration yourself.
Git Viewspace takes a different approach: independent lightweight working directories (views) backed by one shared Git repository.
Each agent gets its own view of a specific commit, using virtual file system drivers, one of which creates a private writable layer for its changes. Unchanged files are read directly from Git through a shared filesystem backend, rather than copied into every workspace. A coordinating daemon manages view lifecycles and serializes publication operations, so agents don’t race against the same Git index. A view takes under 0.5 seconds to create, and uses up minimal disk space even for large repositories. You can have literally hundreds of these with minimal overhead.
The result is a foundation for parallel agent workflows with explicit boundaries:
- Independent edits and branches for each agent.
- Shared repository storage and decoded-content caching.
- Explicit commit, merge, and snapshot-update operations.
- Machine-readable status and command results.
- Recovery handling for interrupted filesystem operations.
- Support for explicitly prepared, pinned submodules.
Git Viewspace is not an agent framework, task scheduler, or security sandbox. It doesn’t decide which agent should do what, and it doesn’t make conflicting changes disappear. It provides the filesystem and Git coordination layer underneath those decisions.
The implementation uses Python, Linux FUSE, and OverlayFS. Version 0.1.0 ships as one installable package, with documentation, regression tests, and real-mount lifecycle and recovery checks.
Licensed under Apache-2.0 through Adapt Framework Solutions.
If you’re building systems where multiple agents collaborate on a repository, I’d welcome feedback—particularly on integration workflows, operational simplicity, and what you’d need before adopting it yourself.
Today I'm making a contribution to the AI-agent-focused open source community: git-viewspace, a way for multiple agents to share the same git repository cheaply. (Link in comments)
When several AI agents work on the same codebase, giving each one a separate clone is straightforward—but it duplicates data and leaves you coordinating branches, commits, and integration yourself.
Git Viewspace takes a different approach: independent lightweight working directories (views) backed by one shared Git repository.
Each agent gets its own view of a specific commit, using virtual file system drivers, one of which creates a private writable layer for its changes. Unchanged files are read directly from Git through a shared filesystem backend, rather than copied into every workspace. A coordinating daemon manages view lifecycles and serializes publication operations, so agents don’t race against the same Git index. A view takes under 0.5 seconds to create, and uses up minimal disk space even for large repositories. You can have literally hundreds of these with minimal overhead.
The result is a foundation for parallel agent workflows with explicit boundaries:
- Independent edits and branches for each agent.
- Shared repository storage and decoded-content caching.
- Explicit commit, merge, and snapshot-update operations.
- Machine-readable status and command results.
- Recovery handling for interrupted filesystem operations.
- Support for explicitly prepared, pinned submodules.
Git Viewspace is not an agent framework, task scheduler, or security sandbox. It doesn’t decide which agent should do what, and it doesn’t make conflicting changes disappear. It provides the filesystem and Git coordination layer underneath those decisions.
The implementation uses Python, Linux FUSE, and OverlayFS. Version 0.1.0 ships as one installable package, with documentation, regression tests, and real-mount lifecycle and recovery checks.
Licensed under Apache-2.0 through Adapt Framework Solutions.
If you’re building systems where multiple agents collaborate on a repository, I’d welcome feedback—particularly on integration workflows, operational simplicity, and what you’d need before adopting it yourself.
Link to repository: github.com/adapt-toolkit/git…
Aleksandr Bulkin retweeted
This is Tommy. Tommy had a question about my allegiances. Listen to the end to hear about his.
English: what are your pronouns?
Russian: what are your pronouns and 63 modifier endings, as well as six past tense verb endings?
Hungarian: We have no gendered constructs whatsoever. If this somehow motivates you, good luck and Godspeed.
Aleksandr Bulkin retweeted
Personally? Oh man.
This is just a tweak on Peter’s question which most of you know:
“What is something you feel is good and needs to be done, but on which almost no institution agrees with you and will fight to destroy you with their full force at a deeply personal level?”
Where to begin??
Undermining faith in Anthony Fauci’s and Francis Collins’ scientific behavior during a pandemic.
Calling attention to the fact that the president of the United States with launch code responsibility was obviously demented during his presidency.
Opposing high levels of immigration for completley non-xenophobic reasons.
Openly discussing asymmetries in patterns of religious martyrdom attacks.
Fighting the loss of the word “Genocide” via coordinated social campaigns.
Pointing out that the history of Peer Review is retconned and fictitious and obviously false.
Fighting irreversible surgeries pushed on children leading to a reproductive holocaust.
Calling for an end to Diverisity-Equity-Inclusion and opposing “Black Lives Matter”, despite their branding genius, even during the Summer of Floyd.
Pointing out that conspiracies are well documented and ubiquitous and fighting for responsible conspiracy theorizing.
Promoting my own technical work and insisting that scientists get paid handsomely.
Etc. Etc.
You’re asking the wrong person. It seems I stand up for what I believe is deeply good and what is misperceived as evil all day, every day. This has stupidly become close to my whole life at times. I’m so bored by how silent everyone is. I can’t fathom it. “Managed Reality” has turned evil.
Something is pushing pure evil, disguised as “good” down our throats 24/7 via the institutions. And almost everyone puts up with it. I do not.
But: Thanks for the question.
A great one!
It’s simple. If the “greatest living mathematician” says obvious nonsense, it’s still nonsense.
It would be fascinating to know where this rumor among non-mathematicians got started about the greatest living mathematician.
Jean Pierre Serre is still alive. As is Edward Witten. And Simon Donaldson. As well as Peter Scholze. And Jacob Lurie. Not to mention Graeme Segal. Or Nigel Hitchin. Or Pierre Deligne. To say nothing of Dan Freed. Or Michael Hopkins. Or Maxim Konsevitch. Or ST Yau and Mikhail Gromov. Or Drinfeld and Beilinson. Or….
I think Terry Tao is amazing.
I cannot shake the feeling that the New York Times and some people at Princeton came up with a meme that Tao won the prize as the greatest living mathematician. Which simply never happened.
I’m sorry, but this is wrong. Nothing against Tao. But It’s just a meme.
It is remarkable how Dubbya is suddenly being remembered by progressives as some kind of a hero of American democracy…
#trumptimes
Isn’t it ironic that the AI doomerism movement and the AI industry as a whole serve as a striking demonstration of the failure mode of intelligence?
A common response is “if they truly believe this, why are they still building it?” At OpenAI, many have not deeply internalized the civilizational stakes. At Anthropic, the stakes are well-understood, but they are locked in a race to get there first - they believe no one else will act responsibly, so they must do it themselves, despite the risk.
Gosh. There’s one person who should probably stay out of the whole programming language safety discussion…
The creator of the C++ programming language, Bjarne Stroustrup, says AI-generated code is not successful and calls the idea of using natural language as a programming language idiotic.
He says humans will still write code and use abstraction.
According to him, AI generates bloated code with more bugs and security holes, making it hard to validate.
He also says the senior developers needed to validate it are starting to retire because they don’t want to deal with validating something that changes every time you change your prompt.
He goes on to say that AI is not good at writing safety-critical or performance-critical code: “Now, let’s say that 70 or 80% of the world’s code doesn’t fit that pattern, but it’s that 10, 20% of the code that I’m interested in.”
Aleksandr Bulkin retweeted
this is maybe a shade too provocative a framing, but ... the wild thing about folks in and close to govt talking about banning foreign tech and taking direct shares in frontier labs to protect them from short-term profit squeezes on the way to global domination is that you can almost frame a question being posed here as something like:
"can the US government become China faster than Kimi can become Claude?"