Search

🔗 https://euvd.enisa.europa.eu/vulnerability/EUVD-2026-28261

EUVD Bot @[email protected] · 2026-05-07 · 10:48 UTC

🚨 EUVD-2026-28261

📊 Score: 7.5/10 (CVSS v3.1)
📦 Product: nocobase
🏢 Vendor: nocobase
📅 Updated: 2026-05-07

📝 NocoBase is an AI-powered no-code/low-code platform for building business applications and enterprise solutions. Prior to version 2.0.39, the queryParentSQL() function in the core database package constructs a recursive CTE query by joining nodeIds with st...

🔗 https://euvd.enisa.europa.eu/vulnerability/EUVD-2026-28261

EUVD Bot @[email protected] · 2026-05-07 · 10:48 UTC

🚨 EUVD-2026-28261

📊 Score: 7.5/10 (CVSS v3.1)
📦 Product: nocobase
🏢 Vendor: nocobase
📅 Updated: 2026-05-07

📝 NocoBase is an AI-powered no-code/low-code platform for building business applications and enterprise solutions. Prior to version 2.0.39, the queryParentSQL() function in the core database package constructs a recursive CTE query by joining nodeIds with st...

🔗 https://euvd.enisa.europa.eu/vulnerability/EUVD-2026-28261

#vulnerability #cve #euvd #infosec #cybersecurity

EUVD Bot @[email protected] · 2026-05-07 · 10:48 UTC

🚨 EUVD-2026-28261

📊 Score: 7.5/10 (CVSS v3.1)
📦 Product: nocobase
🏢 Vendor: nocobase
📅 Updated: 2026-05-07

📝 NocoBase is an AI-powered no-code/low-code platform for building business applications and enterprise solutions. Prior to version 2.0.39, the queryParentSQL() function in the core database package constructs a recursive CTE query by joining nodeIds with st...

Habr @[email protected] · 2026-05-04 · 04:22 UTC

Сепаратор для логов. Сжимаем логи для контекста LLM без потери читаемости

logzip — утилита для сжатия логов перед анализом в LLM. Вместо того чтобы отправлять в Claude/GPT сырые 10 МБ лога, мы сжимаем его до 3.4 МБ (−58%), сохраняя полную читаемость и видимость ошибок. Статья рассказывает о проблеме (Lost in the Middle, переплата за мусорные токены), архитектуре на Rust, алгоритме Recursive BPE и реальном ROI: −$2,070 в год на API. Проект open source (MIT), интегрирован с Claude Desktop через MCP.

#logzip #llmинструменты #mcptools #token_management #compression #log #python #rust #devops #optimization

#optimization #devops #rust #python #log #compression

Habr @[email protected] · 2026-05-04 · 04:22 UTC

Сепаратор для логов. Сжимаем логи для контекста LLM без потери читаемости

logzip — утилита для сжатия логов перед анализом в LLM. Вместо того чтобы отправлять в Claude/GPT сырые 10 МБ лога, мы сжимаем его до 3.4 МБ (−58%), сохраняя полную читаемость и видимость ошибок. Статья рассказывает о проблеме (Lost in the Middle, переплата за мусорные токены), архитектуре на Rust, алгоритме Recursive BPE и реальном ROI: −$2,070 в год на API. Проект open source (MIT), интегрирован с Claude Desktop через MCP.

#logzip #llmинструменты #mcptools #token_management #compression #log #python #rust #devops #optimization

#optimization #devops #rust #python #log #compression

Habr @[email protected] · 2026-05-04 · 04:22 UTC

Сепаратор для логов. Сжимаем логи для контекста LLM без потери читаемости

logzip — утилита для сжатия логов перед анализом в LLM. Вместо того чтобы отправлять в Claude/GPT сырые 10 МБ лога, мы сжимаем его до 3.4 МБ (−58%), сохраняя полную читаемость и видимость ошибок. Статья рассказывает о проблеме (Lost in the Middle, переплата за мусорные токены), архитектуре на Rust, алгоритме Recursive BPE и реальном ROI: −$2,070 в год на API. Проект open source (MIT), интегрирован с Claude Desktop через MCP.

#logzip #llmинструменты #mcptools #token_management #compression #log #python #rust #devops #optimization

#optimization #devops #rust #python #log #compression

Habr @[email protected] · 2026-05-04 · 04:22 UTC

Сепаратор для логов. Сжимаем логи для контекста LLM без потери читаемости

logzip — утилита для сжатия логов перед анализом в LLM. Вместо того чтобы отправлять в Claude/GPT сырые 10 МБ лога, мы сжимаем его до 3.4 МБ (−58%), сохраняя полную читаемость и видимость ошибок. Статья рассказывает о проблеме (Lost in the Middle, переплата за мусорные токены), архитектуре на Rust, алгоритме Recursive BPE и реальном ROI: −$2,070 в год на API. Проект open source (MIT), интегрирован с Claude Desktop через MCP.

#logzip #llmинструменты #mcptools #token_management #compression #log #python #rust #devops #optimization

#logzip #llmинструменты #mcptools #token_management #compression #log

earthling @[email protected] · 2026-04-26 · 09:37 UTC

5/
• From Bacteria to Bach: Tracking the lineage of design from the earliest "Darwinian creatures" to the cultural and memetic evolution that allows for human creativity and the composition of complex works like those of J.S. Bach.

The content provides an overview of how the "Great Divide" between mindless matter and mindful consciousness is bridged through the slow, recursive process of natural selection.

https://youtu.be/CZELb41aM6k

#DanielDennett
#mind
#consciousness
#philosophy
#podcast

#danieldennett #mind #consciousness #philosophy #podcast

TinDrum @[email protected] · 2026-04-20 · 04:13 UTC

RE: https://scholar.social/@Aepasek/116432864586994103

This's pretty good – Anne Pasek

'Some of my colleagues have begun to speak in quasi-geological terms of a historical boundary condition created by the introduction of generative text and image outputs to the web. There is a before and an after to a world with trash AI, and this peculiarly post-anthropocenic informational layer is contaminated with generative slop. This informational pollution is a problem for large language models themselves, with pre-AI web data privileged in some circles as a way to avoid recursive collapses'

#genAI #technology #research #academia #slop #mediaStudies

#genai #technology #research #academia #slop #mediastudies

TinDrum @[email protected] · 2026-04-20 · 04:13 UTC

RE: https://scholar.social/@Aepasek/116432864586994103

This's pretty good – Anne Pasek

'Some of my colleagues have begun to speak in quasi-geological terms of a historical boundary condition created by the introduction of generative text and image outputs to the web. There is a before and an after to a world with trash AI, and this peculiarly post-anthropocenic informational layer is contaminated with generative slop. This informational pollution is a problem for large language models themselves, with pre-AI web data privileged in some circles as a way to avoid recursive collapses'

#genAI #technology #research #academia #slop #mediaStudies

#genai #technology #research #academia #slop #mediastudies

TinDrum @[email protected] · 2026-04-20 · 04:13 UTC

RE: https://scholar.social/@Aepasek/116432864586994103

This's pretty good – Anne Pasek

'Some of my colleagues have begun to speak in quasi-geological terms of a historical boundary condition created by the introduction of generative text and image outputs to the web. There is a before and an after to a world with trash AI, and this peculiarly post-anthropocenic informational layer is contaminated with generative slop. This informational pollution is a problem for large language models themselves, with pre-AI web data privileged in some circles as a way to avoid recursive collapses'

#genAI #technology #research #academia #slop #mediaStudies

#genai #technology #research #academia #slop #mediastudies

TinDrum @[email protected] · 2026-04-20 · 04:13 UTC

RE: https://scholar.social/@Aepasek/116432864586994103

This's pretty good – Anne Pasek

'Some of my colleagues have begun to speak in quasi-geological terms of a historical boundary condition created by the introduction of generative text and image outputs to the web. There is a before and an after to a world with trash AI, and this peculiarly post-anthropocenic informational layer is contaminated with generative slop. This informational pollution is a problem for large language models themselves, with pre-AI web data privileged in some circles as a way to avoid recursive collapses'

#genAI #technology #research #academia #slop #mediaStudies

#mediastudies #slop #academia #research #technology #genai

TinDrum @[email protected] · 2026-04-20 · 04:13 UTC

RE: https://scholar.social/@Aepasek/116432864586994103

This's pretty good – Anne Pasek

'Some of my colleagues have begun to speak in quasi-geological terms of a historical boundary condition created by the introduction of generative text and image outputs to the web. There is a before and an after to a world with trash AI, and this peculiarly post-anthropocenic informational layer is contaminated with generative slop. This informational pollution is a problem for large language models themselves, with pre-AI web data privileged in some circles as a way to avoid recursive collapses'

#genAI #technology #research #academia #slop #mediaStudies

#mediastudies #genai #technology #research #academia #slop

Jonathan Lamothe @[email protected] · 2026-04-12 · 00:44 UTC

CW: elisp help

Okay, I need to do a hacky #elisp thing. Yes, I know it's terrible.

Basically, I have an existing defun. Let's call it foo. I need to replace it with a new function that calls the old one and transforms its output before returning it.

I naïvely assumed I could do it like this:

(let ((oldfunc (function foo)))
  (defun foo ()
    (my-transform (funcall oldfunc))))

...but this doesn't actually copy the old function, just a reference to the symbol, so it ends up locking itself in a recursive loop.

I'm sure there's a way to do this.
#AskFedi

Edit: Got it. It's:

(let ((oldfunc (symbol-function 'foo)))
  (defun foo ()
    (my-transform (funcall oldfunc))))

Edit 2: It turns out there's a cleaner way still.
See: aus.social/@carlozancanaro/116…

Also, there's still something Gmail isn't liking. Looking at the differences in the headers between emacs and my other clients (whose mail does get through), the next most obvious difference is that the Content-Type header doesn't specify an encoding. Whether this is the actual problem or not, I should probably fix that. I'm just working on how.
RE: aus.social/users/carlozancanar…

#askfedi #elisp

chmod777 @[email protected] · 2026-04-08 · 02:04 UTC

Writing my first bottom up parser. I want my xml lexer to give the doctype as one token but to do that I need to parse the internal subset -> markup decl -> element decl -> content spec -> children which has

[47] children ::= (choice | seq) ('?' | '*' | '+')?
[48] cp ::= (Name | choice | seq) ('?' | '*' | '+')?
[49] choice ::= '(' S? cp ( S? '|' S? cp )+ S? ')'
[50] seq ::= '(' S? cp ( S? ',' S? cp )* S? ')'

As its grammer. Notice the recursion. I would normally use a recursive decent parser but since I'm using Rust's coroutines I can't have recursive coroutines(as far as I am aware).

I'm using coroutines because this is a streaming parser meant for embedded systems with very little memory. At any point I could run out of input which is when I yield back up to get more. My previous iteration of this was a massive state machine essentially implementing coroutines from scratch.

chmod777 @[email protected] · 2026-04-08 · 02:04 UTC

Writing my first bottom up parser. I want my xml lexer to give the doctype as one token but to do that I need to parse the internal subset -> markup decl -> element decl -> content spec -> children which has

[47] children ::= (choice | seq) ('?' | '*' | '+')?
[48] cp ::= (Name | choice | seq) ('?' | '*' | '+')?
[49] choice ::= '(' S? cp ( S? '|' S? cp )+ S? ')'
[50] seq ::= '(' S? cp ( S? ',' S? cp )* S? ')'

As its grammer. Notice the recursion. I would normally use a recursive decent parser but since I'm using Rust's coroutines I can't have recursive coroutines(as far as I am aware).

I'm using coroutines because this is a streaming parser meant for embedded systems with very little memory. At any point I could run out of input which is when I yield back up to get more. My previous iteration of this was a massive state machine essentially implementing coroutines from scratch.

chmod777 @[email protected] · 2026-04-08 · 02:04 UTC

Writing my first bottom up parser. I want my xml lexer to give the doctype as one token but to do that I need to parse the internal subset -> markup decl -> element decl -> content spec -> children which has

[47] children ::= (choice | seq) ('?' | '*' | '+')?
[48] cp ::= (Name | choice | seq) ('?' | '*' | '+')?
[49] choice ::= '(' S? cp ( S? '|' S? cp )+ S? ')'
[50] seq ::= '(' S? cp ( S? ',' S? cp )* S? ')'

As its grammer. Notice the recursion. I would normally use a recursive decent parser but since I'm using Rust's coroutines I can't have recursive coroutines(as far as I am aware).

I'm using coroutines because this is a streaming parser meant for embedded systems with very little memory. At any point I could run out of input which is when I yield back up to get more. My previous iteration of this was a massive state machine essentially implementing coroutines from scratch.

chmod777 @[email protected] · 2026-04-08 · 02:04 UTC

Writing my first bottom up parser. I want my xml lexer to give the doctype as one token but to do that I need to parse the internal subset -> markup decl -> element decl -> content spec -> children which has

[47] children ::= (choice | seq) ('?' | '*' | '+')?
[48] cp ::= (Name | choice | seq) ('?' | '*' | '+')?
[49] choice ::= '(' S? cp ( S? '|' S? cp )+ S? ')'
[50] seq ::= '(' S? cp ( S? ',' S? cp )* S? ')'

As its grammer. Notice the recursion. I would normally use a recursive decent parser but since I'm using Rust's coroutines I can't have recursive coroutines(as far as I am aware).

I'm using coroutines because this is a streaming parser meant for embedded systems with very little memory. At any point I could run out of input which is when I yield back up to get more. My previous iteration of this was a massive state machine essentially implementing coroutines from scratch.