# Code

Code is a program's instructions written in a programming language.

This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles.

## Automated Code Efficiency Optimization via Agent Skill Mining

DevFeed: [Automated Code Efficiency Optimization via Agent Skill Mining](<https://devfeed.tech/articles/automated-code-efficiency-optimization-via-agent-skill-mining-137922.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2603.27850>)

Author: Zimu Wang, Yuling Shi, Mengfan Li, Zijun Liu, Jie M. Zhang, Chengcheng Wan, Xiaodong Gu

Published: 2026-11-17T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Software Engineering](<https://devfeed.tech/sources/arxiv-software-engineering.md>)

Topics: [AI-assisted coding](<https://devfeed.tech/topics/ai-assisted-coding.md>), [Code quality](<https://devfeed.tech/topics/code-quality.md>), [Code](<https://devfeed.tech/topics/code.md>), [Test automation](<https://devfeed.tech/topics/test-automation.md>)

Tags: [agent](<https://devfeed.tech/tags/agent.md>), [agent-skill](<https://devfeed.tech/tags/agent-skill.md>), [c-plus-plus](<https://devfeed.tech/tags/c-plus-plus.md>), [candidate](<https://devfeed.tech/tags/candidate.md>), [code](<https://devfeed.tech/tags/code.md>), [cs-cl](<https://devfeed.tech/tags/cs-cl.md>), [cs-se](<https://devfeed.tech/tags/cs-se.md>), [efficiency](<https://devfeed.tech/tags/efficiency.md>), [feedback](<https://devfeed.tech/tags/feedback.md>), [filter](<https://devfeed.tech/tags/filter.md>), [framework](<https://devfeed.tech/tags/framework.md>), [instructions](<https://devfeed.tech/tags/instructions.md>), [language-models](<https://devfeed.tech/tags/language-models.md>), [libraries](<https://devfeed.tech/tags/libraries.md>), [optimization](<https://devfeed.tech/tags/optimization.md>), [pairs](<https://devfeed.tech/tags/pairs.md>), [paper](<https://devfeed.tech/tags/paper.md>), [pipeline](<https://devfeed.tech/tags/pipeline.md>), [skill](<https://devfeed.tech/tags/skill.md>), [tasks](<https://devfeed.tech/tags/tasks.md>), [tests](<https://devfeed.tech/tags/tests.md>), [transformation](<https://devfeed.tech/tags/transformation.md>)

### AI overview

EffiSkill uses persistent Operator Skill cards to guide code optimization: it distills language-specific libraries from slow/fast program pairs, then diagnoses new code, retrieves and composes skills, and generates up to eight candidates. Candidate generation does not use execution feedback; public tests filter and rank candidates, while held-out private tests determine reported correctness and speedup. Across 623 EffiBench-X tasks per language, two model backbones, and separate Python and C++ evaluations, EffiSkill had the highest observed OPT@1 and OPT@8 among compared methods in all four settings. Its OPT@8 lead over the strongest baselines ranged from 4.98 to 12.52 percentage points. The results support broader successful optimization coverage for the complete pipeline under a shared output budget.

### Source excerpt

arXiv:2603.27850v2 Announce Type: replace Abstract: Code efficiency is a fundamental aspect of software quality, yet how to use large language models (LLMs) to optimize correct programs remains challenging. Prior approaches rely on one-shot rewriting, retrieved exemplars or instructions, fine-tuning, or candidate search. In this paper, we present EffiSkill, a two-stage framework that models recurring slow-to-fast transformations as Operator Skills: persistent cards that record a transformation's intent, applicability, procedure, expected effect, and risks. Stage I distills frozen, language-specific skill libraries from slow/fast program pairs. Stage II diagnoses a new program, retrieves relevant skills, composes alternative plans, and generates up to eight candidates. These steps finish without candidate-execution feedback; public tests then filter and rank candidates, and held-out private tests determine reported correctness and speedup. Across 623 EffiBench-X tasks per language, two backbones, and separate Python and C++ evaluations, EffiSkill achieves the highest observed OPT@1 and OPT@8 among the compared methods in all four settings. Its OPT@8 margins over the strongest baselines range from 4.98 to 12.52 percentage points. These results support broader successful optimization coverage for the complete skill-guided pipeline under a shared output budget.

## Does Confusing Code Confuse LLMs? A Pilot Study on C++

DevFeed: [Does Confusing Code Confuse LLMs? A Pilot Study on C++](<https://devfeed.tech/articles/does-confusing-code-confuse-llms-a-pilot-study-on-c-131719.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.15575>)

Author: Tim van de Wetering, Lola Solovyeva, Fernando Castor

Published: 2026-11-13T05:00:00Z

Content type: news

Language: en

Sources: [arXiv: Software Engineering](<https://devfeed.tech/sources/arxiv-software-engineering.md>)

Topics: [Large Language Model](<https://devfeed.tech/topics/llm.md>), [AI-assisted coding](<https://devfeed.tech/topics/ai-assisted-coding.md>), [Code](<https://devfeed.tech/topics/code.md>), [llm-reasoning](<https://devfeed.tech/topics/llm-reasoning.md>)

Tags: [c-plus-plus](<https://devfeed.tech/tags/c-plus-plus.md>), [chatgpt](<https://devfeed.tech/tags/chatgpt.md>), [code](<https://devfeed.tech/tags/code.md>), [copilot](<https://devfeed.tech/tags/copilot.md>), [cs-se](<https://devfeed.tech/tags/cs-se.md>), [github](<https://devfeed.tech/tags/github.md>), [github-copilot](<https://devfeed.tech/tags/github-copilot.md>), [impact](<https://devfeed.tech/tags/impact.md>), [language](<https://devfeed.tech/tags/language.md>), [language-models](<https://devfeed.tech/tags/language-models.md>), [large-language-models](<https://devfeed.tech/tags/large-language-models.md>), [large-language-models-llms](<https://devfeed.tech/tags/large-language-models-llms.md>), [llms](<https://devfeed.tech/tags/llms.md>), [maintenance](<https://devfeed.tech/tags/maintenance.md>), [matching](<https://devfeed.tech/tags/matching.md>), [pattern](<https://devfeed.tech/tags/pattern.md>), [pattern-matching](<https://devfeed.tech/tags/pattern-matching.md>), [patterns](<https://devfeed.tech/tags/patterns.md>), [performance](<https://devfeed.tech/tags/performance.md>), [pilot](<https://devfeed.tech/tags/pilot.md>), [study](<https://devfeed.tech/tags/study.md>)

### AI overview

A pilot study using ChatGPT and GitHub Copilot on C++ code from three production libraries finds that the effect of human-studied "atoms of confusion" depends on the maintenance task. The study tests atom injection and removal, output prediction, and code explanation. Patterns requiring reasoning about surrounding code are more resistant to transformation, while localized syntactic patterns are handled reliably. Both models remove these patterns more reliably than they inject them; the authors suggest removal may rely on pattern matching, whereas injection requires deeper comprehension. These are preliminary findings from a pilot study.

### Source excerpt

arXiv:2611.15575v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly deployed for software maintenance tasks that depend heavily on code comprehension. While atoms of confusion, minimal code patterns empirically shown to impair human program comprehension, are well studied, it remains unclear whether these same patterns affect LLMs, and whether their impact varies across maintenance tasks. We present a pilot study evaluating the impact of atoms of confusion on LLM performance on C++ code drawn from three production libraries, using ChatGPT and GitHub Copilot on three software development tasks: atom injection and removal, output prediction, and code explanation. Our results reveal three preliminary findings: (i) the impact of atoms on LLMs is strongly task-dependent; (ii) atoms requiring reasoning about surrounding code are the most resistant to transformation, while localized syntactic atoms are handled reliably; and (iii) LLMs remove atoms of confusion more reliably than they inject them, suggesting that removal succeeds through pattern matching while injection requires deeper comprehension.

## Program-Aided Patchable Skill Evolution

DevFeed: [Program-Aided Patchable Skill Evolution](<https://devfeed.tech/articles/program-aided-patchable-skill-evolution-128388.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.14019>)

Author: Jiahao Wang, Yuqi Xiong, Qing Yang, Zhipeng Xu, Zhenghao Liu, Zulong Chen, Yubin Bao

Published: 2026-11-11T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Agent Skills](<https://devfeed.tech/topics/agent-skills.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [Code](<https://devfeed.tech/topics/code.md>)

Tags: [agent](<https://devfeed.tech/tags/agent.md>), [agent-skills](<https://devfeed.tech/tags/agent-skills.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cl](<https://devfeed.tech/tags/cs-cl.md>), [distilling](<https://devfeed.tech/tags/distilling.md>), [evolution](<https://devfeed.tech/tags/evolution.md>), [experiments](<https://devfeed.tech/tags/experiments.md>), [feedback](<https://devfeed.tech/tags/feedback.md>), [framework](<https://devfeed.tech/tags/framework.md>), [github](<https://devfeed.tech/tags/github.md>), [improvements](<https://devfeed.tech/tags/improvements.md>), [knowledge](<https://devfeed.tech/tags/knowledge.md>), [llm](<https://devfeed.tech/tags/llm.md>), [models](<https://devfeed.tech/tags/models.md>), [natural-language](<https://devfeed.tech/tags/natural-language.md>), [procedural](<https://devfeed.tech/tags/procedural.md>), [program](<https://devfeed.tech/tags/program.md>), [skills](<https://devfeed.tech/tags/skills.md>), [tasks](<https://devfeed.tech/tags/tasks.md>)

### AI overview

PASE evolves LLM agent skills by pairing them with executable programs. It separates reusable procedural logic into editable code, so that logic can be updated directly instead of repeatedly revising natural language skill documents that mix procedures with descriptive knowledge. The abstract reports that PASE outperforms all evaluated baselines by over 2.8% across diverse models and benchmarks, with greater evolutionary stability and generalization. It does not specify the benchmarks, comparison metrics, or how the framework is implemented.

### Source excerpt

arXiv:2611.14019v1 Announce Type: cross Abstract: Agent skills have emerged as an effective means of enabling LLM agents to solve complex, interactive tasks by encoding reusable knowledge without modifying model parameters. Recent approaches further improve skill adaptation by distilling reusable experience from agent execution trajectories and evolving skills based on downstream task feedback. However, existing methods primarily represent skills as natural language documents, intertwining procedural logic with descriptive knowledge and making skills difficult to maintain and evolve. To address this limitation, we propose PASE (Program-Aided Patchable Skill Evolution), a program-aided framework that iteratively evolves skills by associating them with executable programs. Unlike prior approaches that encode trajectory-derived experience primarily in natural language, PASE separates reusable procedural logic into editable programs, enabling targeted updates to executable components while reducing the need to repeatedly revise natural language skills. Extensive experiments across diverse models and benchmarks demonstrate that PASE consistently outperforms all baselines, by achieving over 2.8% improvements. Further analyses show that PASE exhibits greater evolutionary stability and generalisation capability, highlighting the critical role of programmatic knowledge in skill evolution. We will release all data and code via GitHub.

## CVE-2026-109097 - ROS# Code Injection Vulnerability

DevFeed: [CVE-2026-109097 - ROS# Code Injection Vulnerability](<https://devfeed.tech/articles/cve-2026-109097-ros-code-injection-vulnerability-125785.md>)

Original publisher: [Read original article](<https://cvefeed.io/vuln/detail/CVE-2026-109097>)

Author: CVEFEED.io; siemens.com

Published: 2026-11-10T12:17:16Z

Content type: news

Language: en

Sources: [Latest High/Critical Vulnerabilitiy Feed](<https://devfeed.tech/sources/latest-high-critical-vulnerabilitiy-feed.md>)

Topics: [Code](<https://devfeed.tech/topics/code.md>)

Tags: [2026](<https://devfeed.tech/tags/2026.md>), [code-injection](<https://devfeed.tech/tags/code-injection.md>), [cve](<https://devfeed.tech/tags/cve.md>), [cve-2026-109097](<https://devfeed.tech/tags/cve-2026-109097.md>), [ros](<https://devfeed.tech/tags/ros.md>), [siemens](<https://devfeed.tech/tags/siemens.md>), [vulnerability](<https://devfeed.tech/tags/vulnerability.md>)

### AI overview

ROS# versions before 2.3.1 do not properly escape text in message definition files. An attacker able to provide a crafted definition file could inject arbitrary code into the generated program. The excerpt rates the vulnerability 8.6 (HIGH); it gives no mitigation or remediation details.

### Source excerpt

CVE ID :CVE-2026-109097 Published : Nov. 10, 2026, 12:17 p.m. | 14 minutes ago Description :A vulnerability has been identified in ROS# (All versions < V2.3.1). Affected systems do not properly escape text of message definition files. This could allow an attacker that is able to pass a crafted message definition file to the user to inject arbitrary code into the generated program. Severity: 8.6 | HIGH Visit the link for more details, such as CVSS details, affected products, timeline, and more...

## Introducing AlphaProof Code: a breakthrough in verifying complex software at depth

DevFeed: [Introducing AlphaProof Code: a breakthrough in verifying complex software at depth](<https://devfeed.tech/articles/introducing-alphaproof-code-a-breakthrough-in-verifying-complex-software-at-depth-128945.md>)

Original publisher: [Read original article](<https://deepmind.google/blog/introducing-alphaproof-code/>)

Author: Pushmeet Kohli; Swarat Chaudhuri

Published: 2026-11-10T12:02:28Z

Content type: release

Language: en

Sources: [Google DeepMind News](<https://devfeed.tech/sources/google-deepmind-news.md>)

Topics: [Formal methods](<https://devfeed.tech/topics/formal-methods.md>), [Lean](<https://devfeed.tech/topics/lean.md>), [Multi Agent Systems](<https://devfeed.tech/topics/multi-agent-systems.md>), [code security](<https://devfeed.tech/topics/code-security.md>), [Code](<https://devfeed.tech/topics/code.md>)

Tags: [after](<https://devfeed.tech/tags/after.md>), [agentic](<https://devfeed.tech/tags/agentic.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [all](<https://devfeed.tech/tags/all.md>), [also](<https://devfeed.tech/tags/also.md>), [apps](<https://devfeed.tech/tags/apps.md>), [automatically](<https://devfeed.tech/tags/automatically.md>), [banking](<https://devfeed.tech/tags/banking.md>), [breakthrough](<https://devfeed.tech/tags/breakthrough.md>), [bridging](<https://devfeed.tech/tags/bridging.md>), [code](<https://devfeed.tech/tags/code.md>), [complex](<https://devfeed.tech/tags/complex.md>), [core](<https://devfeed.tech/tags/core.md>), [lean](<https://devfeed.tech/tags/lean.md>), [proofs](<https://devfeed.tech/tags/proofs.md>), [research](<https://devfeed.tech/tags/research.md>), [signal](<https://devfeed.tech/tags/signal.md>), [software](<https://devfeed.tech/tags/software.md>), [verify](<https://devfeed.tech/tags/verify.md>)

### AI overview

Google DeepMind reports that AlphaProof Code formally verified key security properties of Signal's compiled cryptographic code, connecting protocol-level goals to the implementation running on devices. Human experts and a mix of models translated the goals into mathematical specifications, which experts reviewed; Gemini Carbon agents then generated Lean proofs, and agents audited them, flagging about one percent of the codebase for further human review. Lean machine-checked the proofs. The work took four weeks, and in tests with two previously fixed bugs reintroduced into an isolated code version, the system proved the buggy versions insecure. The account describes verification of specified key properties, not a guarantee that all possible flaws are absent.

### Source excerpt

AlphaProof Code agents wrote three million lines of proofs in Lean to verify Signal's core encryption.

## VeriContest: A Competitive-Programming Benchmark for Verifiable Code Generation

DevFeed: [VeriContest: A Competitive-Programming Benchmark for Verifiable Code Generation](<https://devfeed.tech/articles/vericontest-a-competitive-programming-benchmark-for-verifiable-code-generation-127184.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2605.08553>)

Author: Zichen Xie, Mrigank Pawagi, Yuxin Liu, Aaditi Rai, Lize Shao, John Berberian Jr., Sicong Che, Wenxi Wang

Published: 2026-11-10T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Software Engineering](<https://devfeed.tech/sources/arxiv-software-engineering.md>)

Topics: [competitive-programming](<https://devfeed.tech/topics/competitive-programming.md>), [Automated reasoning](<https://devfeed.tech/topics/automated-reasoning.md>), [Code](<https://devfeed.tech/topics/code.md>), [Ground truth / benchmark quality](<https://devfeed.tech/topics/ground-truth-benchmark-quality.md>), [LeetCode](<https://devfeed.tech/topics/leetcode.md>)

Tags: [benchmark](<https://devfeed.tech/tags/benchmark.md>), [bottlenecks](<https://devfeed.tech/tags/bottlenecks.md>), [code-generation](<https://devfeed.tech/tags/code-generation.md>), [competitive-programming](<https://devfeed.tech/tags/competitive-programming.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-se](<https://devfeed.tech/tags/cs-se.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [formal](<https://devfeed.tech/tags/formal.md>), [large-language-models](<https://devfeed.tech/tags/large-language-models.md>), [proofs](<https://devfeed.tech/tags/proofs.md>), [rust](<https://devfeed.tech/tags/rust.md>), [synthesis](<https://devfeed.tech/tags/synthesis.md>), [test](<https://devfeed.tech/tags/test.md>), [testing](<https://devfeed.tech/tags/testing.md>), [verus](<https://devfeed.tech/tags/verus.md>)

### AI overview

VeriContest is a benchmark for evaluating code generation with machine-checkable correctness: it contains 1,006 LeetCode and Codeforces problems in Rust with Verus, including expert-validated specifications, accepted code, checked proofs, and positive and negative tests. Its three-phase construction combines manually verified seeds with semi-automated expansion and human review; testing also checks postcondition completeness. The benchmark supports isolated, compositional, and end-to-end evaluation. Across ten models, the strongest scored 97.51% on natural-language-to-code generation, versus 57.16% for specifications, 13.82% for proofs, and 6.76% end-to-end, indicating that specifications and proofs remain major bottlenecks.

### Source excerpt

arXiv:2605.08553v2 Announce Type: replace Abstract: Large language models can generate useful code from natural language, but their outputs come without correctness guarantees. Verifiable code generation offers a path beyond testing by requiring models to produce not only executable code, but also formal specifications and machine-checkable proofs. Progress in this direction, however, is difficult to measure: existing benchmarks are often small, cover only part of the pipeline, lack ground-truth proofs or rigorous specification validation, or target verification settings far from mainstream software development. We present VeriContest, a benchmark of 1,006 competitive-programming problems from LeetCode and Codeforces for verifiable code generation in Rust with Verus. Each problem includes a natural language description, expert-validated formal specifications, judge-accepted Rust code, Verus-checked proofs, and positive and negative test suites. VeriContest is constructed through a three-phase pipeline that scales from manually verified seed problems to semi-automated expansion with human-in-the-loop review. To strengthen benchmark quality, we also use testing to validate postcondition completeness. VeriContest supports both isolated and compositional evaluation of specification, code, and proof generation, and end-to-end verified program synthesis. Evaluating ten state-of-the-art models reveals a sharp gap between ordinary coding and verifiable code generation: the strongest model reaches 97.51% on natural-language-to-code generation, but only 57.16% on specification generation, 13.82% on proof generation, and 6.76% end-to-end. These results identify proof and specification generation as the central bottlenecks for current models and establish VeriContest as a rigorous platform for evaluating and training future systems that generate code with machine-checkable correctness.

## Safe by Accident: Crash-Testing LLM-Written Code

DevFeed: [Safe by Accident: Crash-Testing LLM-Written Code](<https://devfeed.tech/articles/safe-by-accident-crash-testing-llm-written-code-127237.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.10585>)

Author: Afridi Mahmud, Md. As-Aid Rahman, Rifat Shahriyar

Published: 2026-11-10T05:00:00Z

Content type: news

Language: en

Sources: [arXiv: Software Engineering](<https://devfeed.tech/sources/arxiv-software-engineering.md>)

Topics: [Transactions](<https://devfeed.tech/topics/transactions.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [Test automation](<https://devfeed.tech/topics/test-automation.md>), [Code](<https://devfeed.tech/topics/code.md>), [AI-assisted coding](<https://devfeed.tech/topics/ai-assisted-coding.md>)

Tags: [api](<https://devfeed.tech/tags/api.md>), [backend](<https://devfeed.tech/tags/backend.md>), [code](<https://devfeed.tech/tags/code.md>), [configuration](<https://devfeed.tech/tags/configuration.md>), [crash](<https://devfeed.tech/tags/crash.md>), [cs-se](<https://devfeed.tech/tags/cs-se.md>), [database](<https://devfeed.tech/tags/database.md>), [email](<https://devfeed.tech/tags/email.md>), [environment](<https://devfeed.tech/tags/environment.md>), [evaluations](<https://devfeed.tech/tags/evaluations.md>), [file](<https://devfeed.tech/tags/file.md>), [library](<https://devfeed.tech/tags/library.md>), [llm](<https://devfeed.tech/tags/llm.md>), [model](<https://devfeed.tech/tags/model.md>), [python](<https://devfeed.tech/tags/python.md>), [repair](<https://devfeed.tech/tags/repair.md>), [safety](<https://devfeed.tech/tags/safety.md>), [sqlite3](<https://devfeed.tech/tags/sqlite3.md>), [testing](<https://devfeed.tech/tags/testing.md>), [transaction](<https://devfeed.tech/tags/transaction.md>)

### AI overview

The study finds that apparent crash safety in LLM-written database code can depend on sqlite3's default transaction behavior. Researchers injected crashes before database, file, and email calls and evaluated 9,100 solutions from ten models across 24 backend tasks. Of database code that survived every crash under sqlite3's default mode, 54% was left partially updated in autocommit mode; this occurred across all models, including code with conn: blocks that do not open transactions. Telling models to use autocommit did not change local code models' output, while API models' unsafe share fell by 30 percentage points. Explicitly asking for crash safety helped every model, and strong models repaired nearly all unsafe code, though they often wrote safe code only when prompted. The abstract suggests testing under a less forgiving configuration to distinguish code-designed safety from protection supplied by the library.

### Source excerpt

arXiv:2611.10585v1 Announce Type: new Abstract: Evaluations of LLM-written code run it once, in one environment, and credit the code with whatever passes. We show that this credit can belong to the environment. We crashed LLM-written backend code just before each database, file and email call, re-ran it as an operator would, and judged the result under two transaction modes of Python's sqlite3: the default, in which the library quietly opens a transaction, and autocommit, in which it does not. Across 9,100 solutions from ten models on 24 backend tasks, more than half (54%) of the database code that survives every crash in the default mode is left half-updated in autocommit mode. This holds for every model, and it hides in ordinary code, including with conn: blocks that look like transactions but open none. Telling the model that its connection is in autocommit mode does not change what the local code models write, while the API models' unsafe share falls by 30 points. Asking for crash safety helps every model, and strong models repair nearly all unsafe code, but often write safe code only when asked. Thus, much of the crash safety that LLM-written database code appears to have is borrowed from the library rather than designed into it, and a second test under a less forgiving configuration reveals the difference.

## Viverra: Text-to-Code with Guarantees

DevFeed: [Viverra: Text-to-Code with Guarantees](<https://devfeed.tech/articles/viverra-text-to-code-with-guarantees-127188.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2605.14972>)

Author: Haoze Wu, Rocky Klopfenstein, Keith Farkas, Nina Narodytska

Published: 2026-11-10T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Software Engineering](<https://devfeed.tech/sources/arxiv-software-engineering.md>)

Topics: [Generative AI](<https://devfeed.tech/topics/generative-ai.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [generative-ai-tools](<https://devfeed.tech/topics/generative-ai-tools.md>), [Code](<https://devfeed.tech/topics/code.md>), [Parsing](<https://devfeed.tech/topics/parsing.md>), [Test-driven development](<https://devfeed.tech/topics/tdd.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-coding-tools](<https://devfeed.tech/tags/ai-coding-tools.md>), [annotations](<https://devfeed.tech/tags/annotations.md>), [assertions](<https://devfeed.tech/tags/assertions.md>), [automatically](<https://devfeed.tech/tags/automatically.md>), [bounded](<https://devfeed.tech/tags/bounded.md>), [c](<https://devfeed.tech/tags/c.md>), [code](<https://devfeed.tech/tags/code.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-hc](<https://devfeed.tech/tags/cs-hc.md>), [cs-lo](<https://devfeed.tech/tags/cs-lo.md>), [cs-se](<https://devfeed.tech/tags/cs-se.md>), [efficiently](<https://devfeed.tech/tags/efficiently.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [llm](<https://devfeed.tech/tags/llm.md>), [natural-language](<https://devfeed.tech/tags/natural-language.md>), [safety](<https://devfeed.tech/tags/safety.md>), [study](<https://devfeed.tech/tags/study.md>), [system](<https://devfeed.tech/tags/system.md>), [tasks](<https://devfeed.tech/tags/tasks.md>), [users](<https://devfeed.tech/tags/users.md>)

### AI overview

Viverra pairs LLM-generated C programs with candidate assertions about safety and correctness, then checks those assertions compositionally and on a best-effort basis using a portfolio of bounded model checkers. The verified annotations are intended to help developers understand generated code; they do not remove the stated need for developers to review, test, and maintain it. Evaluation across 17 programming tasks found the system could efficiently produce code with verified assertions, and a user study with more than 400 task completions found the assertions improved code-comprehension performance.

### Source excerpt

arXiv:2605.14972v2 Announce Type: replace Abstract: A fundamental limitation of Text-to-Code is that no guarantee can be obtained about the correctness of the generated code. Therefore, to ensure its correctness, the generated code still has to be reviewed, tested, and maintained by developers. However, parsing through LLM-generated code can be tedious and time-consuming, potentially negating the productivity gains promised by AI-coding tools. To address this challenge, we present Viverra, a system that automatically produces formally verified annotations alongside generated code to aid user's understanding of the generated program. Given a natural-language task description, Viverra prompts an LLM to synthesize a C program together with candidate assertions expressing safety and correctness properties. It then verifies those assertions in a compositional and best-effort manner via a portfolio of bounded model checkers. Evaluation on 17 diverse programming tasks suggests that Viverra can efficiently generate code with verified assertions, and that these assertions improve users' performance on code-comprehension tasks in a user study with more than 400 task completions.

## Analyzing Code Provenance in LLMs for Programming Education

DevFeed: [Analyzing Code Provenance in LLMs for Programming Education](<https://devfeed.tech/articles/analyzing-code-provenance-in-llms-for-programming-education-124024.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.09177>)

Author: Subhankar Maity, Faiza Belbachir, Omar Boucelma, Guilherme Medeiros Machado

Published: 2026-11-09T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Software Engineering](<https://devfeed.tech/sources/arxiv-software-engineering.md>)

Topics: [LLM evaluation / benchmarking](<https://devfeed.tech/topics/llm-evaluation-benchmarking.md>), [Code](<https://devfeed.tech/topics/code.md>), [Python](<https://devfeed.tech/topics/python.md>), [Stack Overflow](<https://devfeed.tech/topics/stackoverflow.md>)

Tags: [assessment](<https://devfeed.tech/tags/assessment.md>), [code](<https://devfeed.tech/tags/code.md>), [code-generation](<https://devfeed.tech/tags/code-generation.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-se](<https://devfeed.tech/tags/cs-se.md>), [deepseek](<https://devfeed.tech/tags/deepseek.md>), [detection](<https://devfeed.tech/tags/detection.md>), [education](<https://devfeed.tech/tags/education.md>), [gemini](<https://devfeed.tech/tags/gemini.md>), [gpt](<https://devfeed.tech/tags/gpt.md>), [gpt-5](<https://devfeed.tech/tags/gpt-5.md>), [higher-education](<https://devfeed.tech/tags/higher-education.md>), [inference](<https://devfeed.tech/tags/inference.md>), [language](<https://devfeed.tech/tags/language.md>), [language-models](<https://devfeed.tech/tags/language-models.md>), [large-language-models](<https://devfeed.tech/tags/large-language-models.md>), [large-language-models-llms](<https://devfeed.tech/tags/large-language-models-llms.md>), [llama](<https://devfeed.tech/tags/llama.md>), [llama-3-3](<https://devfeed.tech/tags/llama-3-3.md>), [llms](<https://devfeed.tech/tags/llms.md>), [programming](<https://devfeed.tech/tags/programming.md>), [provenance](<https://devfeed.tech/tags/provenance.md>)

### AI overview

The study extends the Min-K% probability method with synthetic non-member code generation to infer whether Python code was memorized by a model and help trace its provenance. It evaluates GPT-2, DistilGPT-2, and DialoGPT alongside GPT-5, Gemini 2.5, Mistral, DeepSeek, and LLaMA 3.3, first checking feasibility with Stack Overflow data before planned evaluation on higher-education assignments. The abstract reports more sensitive code-memorization detection with free models than commercial models and frames the comparison as a trade-off between detection accuracy and privacy. It presents provenance detection together with process visibility as an initial approach to assessing AI-assisted programming education.

### Source excerpt

arXiv:2611.09177v1 Announce Type: cross Abstract: Large language models (LLMs) play an increasing role in programming education, challenging traditional assessment methods as they can generate fully functional code instantly. This disrupts the reliability of student submissions as indicators of learning and effort. To address this issue, code provenance is essential for tracing the origins of code and distinguishing student-authored, AI-generated, and externally sourced components. We extend the Min-K% probability approach by incorporating synthetic non-member code generation and apply it to Python code for membership inference and provenance tracking. Evaluations cover free LLMs such as GPT-2, DistilGPT-2, and DialoGPT, as well as state-of-the-art commercial models including GPT-5, Gemini 2.5, Mistral, DeepSeek, and LLaMA 3.3. Before evaluating our approach on higher-education student assignments, we validate its feasibility using Stack Overflow data. The results show that free models enable more sensitive code memorization detection than commercial models. By systematically comparing these LLMs, we assess trade-offs between detection accuracy and privacy protection, taking an initial step toward effective assessment of AI-assisted programming education through combined provenance detection and process visibility.

## Prompt Less, Smile More: Semantic Engineering in Lieu of Prompt Engineering

DevFeed: [Prompt Less, Smile More: Semantic Engineering in Lieu of Prompt Engineering](<https://devfeed.tech/articles/prompt-less-smile-more-semantic-engineering-in-lieu-of-prompt-engineering-113459.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2511.19427>)

Author: Jayanaka L. Dantanarayana, Savini Kashmira, Xiaoyu Ma, Krisztian Flautner, Lingjia Tang, Jason Mars

Published: 2026-11-03T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Software Engineering](<https://devfeed.tech/sources/arxiv-software-engineering.md>)

Topics: [Prompt Engineering](<https://devfeed.tech/topics/prompt-engineering.md>), [LLM evaluation / benchmarking](<https://devfeed.tech/topics/llm-evaluation-benchmarking.md>), [Code](<https://devfeed.tech/topics/code.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-integrated](<https://devfeed.tech/tags/ai-integrated.md>), [annotations](<https://devfeed.tech/tags/annotations.md>), [applications](<https://devfeed.tech/tags/applications.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-se](<https://devfeed.tech/tags/cs-se.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [language-models](<https://devfeed.tech/tags/language-models.md>), [large-language-models-llms](<https://devfeed.tech/tags/large-language-models-llms.md>), [llms](<https://devfeed.tech/tags/llms.md>), [methodology](<https://devfeed.tech/tags/methodology.md>), [mtp](<https://devfeed.tech/tags/mtp.md>), [performance](<https://devfeed.tech/tags/performance.md>), [prompt](<https://devfeed.tech/tags/prompt.md>), [prompt-engineering](<https://devfeed.tech/tags/prompt-engineering.md>), [semantic](<https://devfeed.tech/tags/semantic.md>)

### AI overview

Meaning Typed Programming (MTP) builds prompts from program structure, but it may leave selection criteria, output requirements, and application policies unspecified. Semantic Engineering addresses those gaps by adding and refining guidance through SemTexts: natural-language instructions attached to program entities, with declaration scopes controlling where they apply. An implementation in the Jac compiler and MTP runtime selects applicable guidance during automatic prompt construction. In an evaluation spanning seven AI-integrated applications, two benchmarks, and three models, it reduced integration lines of code 5.46-fold relative to Prompt Engineering and achieved equal or better task performance in most model-application settings, with a reported maximum of 1.83 times Prompt Engineering's performance.

### Source excerpt

arXiv:2511.19427v2 Announce Type: replace Abstract: AI-integrated applications increasingly rely on large language models (LLMs), but their intended behavior is not always fully specified by code. Meaning Typed Programming (MTP) automates prompt construction from program structure, yet selection criteria, output requirements, and application policies can remain unstated. We introduce Semantic Engineering, a methodology for identifying such missing intent and selectively adding and refining model guidance. We realize this methodology through Semantic Context Annotations (SemTexts), a dedicated language mechanism independent of documentation and schema metadata. SemTexts associate natural-language instructions with individual program entities and use declaration scopes to determine where they apply. Our implementation extends the Jac compiler and MTP runtime to preserve these associations and select applicable guidance during automatic prompt construction. We evaluate Semantic Engineering on seven AI-integrated applications and two standard benchmarks across three models. Across the applications, it achieves a 5.46-fold reduction in integration lines of code relative to Prompt Engineering while delivering equal or better task performance in most model-application settings, reaching up to 1.83 times the task performance of Prompt Engineering.

## Documenting Python Code: A Complete Guide

DevFeed: [Documenting Python Code: A Complete Guide](<https://devfeed.tech/articles/documenting-python-code-a-complete-guide-110726.md>)

Original publisher: [Read original article](<https://realpython.com/documenting-python-code/>)

Author: James Mertz

Published: 2026-11-01T14:00:00Z

Content type: tutorial

Language: en

Sources: [Real Python](<https://devfeed.tech/sources/real-python.md>)

Topics: [Python](<https://devfeed.tech/topics/python.md>), [Tutorial](<https://devfeed.tech/topics/tutorial.md>), [code comments](<https://devfeed.tech/topics/code-comments.md>), [Code](<https://devfeed.tech/topics/code.md>)

Tags: [2](<https://devfeed.tech/tags/2.md>), [2017](<https://devfeed.tech/tags/2017.md>), [3](<https://devfeed.tech/tags/3.md>), [about](<https://devfeed.tech/tags/about.md>), [added](<https://devfeed.tech/tags/added.md>), [alert](<https://devfeed.tech/tags/alert.md>), [algorithm](<https://devfeed.tech/tags/algorithm.md>), [algorithms](<https://devfeed.tech/tags/algorithms.md>), [code](<https://devfeed.tech/tags/code.md>), [complete](<https://devfeed.tech/tags/complete.md>), [guide](<https://devfeed.tech/tags/guide.md>), [know](<https://devfeed.tech/tags/know.md>), [large](<https://devfeed.tech/tags/large.md>), [need](<https://devfeed.tech/tags/need.md>), [project](<https://devfeed.tech/tags/project.md>), [python](<https://devfeed.tech/tags/python.md>), [script](<https://devfeed.tech/tags/script.md>), [small](<https://devfeed.tech/tags/small.md>), [this](<https://devfeed.tech/tags/this.md>), [you-re](<https://devfeed.tech/tags/you-re.md>)

### AI overview

The guide distinguishes comments, which explain code intent to maintainers, from documentation, which explains use and functionality to users. It recommends keeping comments brief, close to the code, and nonredundant, and using type hints where helpful while recognizing they can add documentation work. It explains how docstrings populate an object's `__doc__` and appear through `help()`, and outlines concise one-line or structured multi-line docstrings for classes, methods, modules, packages, functions, and scripts. Suggested content includes purpose, parameters, return values, attributes, side effects, exceptions, and usage requirements. It presents Google, reStructuredText, NumPy/SciPy, and Epytext formats, advising consistency within a project. Project guidance scales with audience: private projects may need a README and examples; shared projects benefit from contribution guidance; public projects may also need a code of conduct, license, and fuller docs organized as tutorials, how-to guides, references, and explanations.

### Source excerpt

A complete guide to documenting Python code. Whether you're documenting a small script or a large project, this guide covers what you need to know.

## Malicious Claude Code extension on Cursor planted SSH backdoors

DevFeed: [Malicious Claude Code extension on Cursor planted SSH backdoors](<https://devfeed.tech/articles/malicious-claude-code-extension-on-cursor-planted-ssh-backdoors-108390.md>)

Original publisher: [Read original article](<https://www.bleepingcomputer.com/news/security/malicious-claude-code-extension-on-cursor-planted-ssh-backdoors/>)

Author: Bill Toulas

Published: 2026-10-30T14:36:54Z

Content type: news

Language: en

Sources: [BleepingComputer](<https://devfeed.tech/sources/bleepingcomputer.md>)

Topics: [Claude Code](<https://devfeed.tech/topics/claude-code.md>), [VS Code Extension](<https://devfeed.tech/topics/vscode-extension.md>), [malicious packages](<https://devfeed.tech/topics/malicious-packages.md>), [ssh](<https://devfeed.tech/topics/ssh.md>), [OpenSSH](<https://devfeed.tech/topics/openssh.md>), [Visual Studio Code](<https://devfeed.tech/topics/visual-studio-code.md>), [Code](<https://devfeed.tech/topics/code.md>), [coding assistant](<https://devfeed.tech/topics/coding-assistant.md>)

Tags: [anthropic](<https://devfeed.tech/tags/anthropic.md>), [backdoor](<https://devfeed.tech/tags/backdoor.md>), [backdoors](<https://devfeed.tech/tags/backdoors.md>), [claude](<https://devfeed.tech/tags/claude.md>), [claude-code](<https://devfeed.tech/tags/claude-code.md>), [computer-help](<https://devfeed.tech/tags/computer-help.md>), [computer-security](<https://devfeed.tech/tags/computer-security.md>), [computers](<https://devfeed.tech/tags/computers.md>), [cursor](<https://devfeed.tech/tags/cursor.md>), [developers](<https://devfeed.tech/tags/developers.md>), [distributed](<https://devfeed.tech/tags/distributed.md>), [extension](<https://devfeed.tech/tags/extension.md>), [extensions](<https://devfeed.tech/tags/extensions.md>), [infosec](<https://devfeed.tech/tags/infosec.md>), [infosec-computer-security](<https://devfeed.tech/tags/infosec-computer-security.md>), [linux](<https://devfeed.tech/tags/linux.md>), [mac](<https://devfeed.tech/tags/mac.md>), [malicious](<https://devfeed.tech/tags/malicious.md>), [malware](<https://devfeed.tech/tags/malware.md>), [malware-removal](<https://devfeed.tech/tags/malware-removal.md>), [marketplace](<https://devfeed.tech/tags/marketplace.md>), [october](<https://devfeed.tech/tags/october.md>), [security](<https://devfeed.tech/tags/security.md>), [spyware](<https://devfeed.tech/tags/spyware.md>), [ssh](<https://devfeed.tech/tags/ssh.md>), [support](<https://devfeed.tech/tags/support.md>), [tech-support](<https://devfeed.tech/tags/tech-support.md>), [technical-support](<https://devfeed.tech/tags/technical-support.md>), [version](<https://devfeed.tech/tags/version.md>), [virus](<https://devfeed.tech/tags/virus.md>), [virus-removal](<https://devfeed.tech/tags/virus-removal.md>), [vs-code](<https://devfeed.tech/tags/vs-code.md>), [vscode](<https://devfeed.tech/tags/vscode.md>), [windows](<https://devfeed.tech/tags/windows.md>)

### AI overview

A compromised Claude Code extension (version 2.1.313) was available only through Cursor's marketplace for about 11 hours on October 29, potentially exposing up to 5,000 developers. A JavaScript loader ran when the extension loaded, collected system details, installed an attacker-controlled SSH key, and set up persistence and a recurring reverse-tunnel attempt. Researchers reported flaws that often prevented SSH access: on macOS, a status check lacked elevated privileges; on Windows, setup was attempted through the unavailable /bin/sh. The malware could still send reconnaissance data, including usernames, hostnames, public IPs, Git email addresses, and names of credential-related environment variables; the excerpt says it did not directly steal credentials. Cursor removed the extension, took its marketplace offline for an audit and key rotation, and worked to disable the attacker's server. Cursor said it found no evidence that other versions or official distribution were affected, and later reporting identified Cursor as the only source. Those who installed it during the exposure window should check for an "svc-telemetry" SSH key and follow Cursor's remediation instructions if they,

### Source excerpt

A malicious version of Anthropic's Claude Code VS Code extension was distributed through Cursor's extension marketplace for approximately 11 hours on October 29, exposing up to 5,000 developers to malware designed to install persistent SSH backdoors. [...]

## ZDI-26-824: (0Day) (Pwn2Own) Alpine iLX-F511 Framework Service Exposed Dangerous Function Remote Code Execution Vulnerability

DevFeed: [ZDI-26-824: (0Day) (Pwn2Own) Alpine iLX-F511 Framework Service Exposed Dangerous Function Remote Code Execution Vulnerability](<https://devfeed.tech/articles/zdi-26-824-0day-pwn2own-alpine-ilx-f511-framework-service-exposed-dangerous-function-remote-code-execution-vulnerability-106020.md>)

Original publisher: [Read original article](<http://www.zerodayinitiative.com/advisories/ZDI-26-824/>)

Published: 2026-10-28T05:00:00Z

Content type: release

Language: en

Sources: [Zero Day Initiative](<https://devfeed.tech/sources/zero-day-initiative.md>)

Topics: [Exploit](<https://devfeed.tech/topics/exploit.md>), [Code](<https://devfeed.tech/topics/code.md>)

Tags: [2026](<https://devfeed.tech/tags/2026.md>), [alpine](<https://devfeed.tech/tags/alpine.md>), [code-execution](<https://devfeed.tech/tags/code-execution.md>), [cve-2026-118606](<https://devfeed.tech/tags/cve-2026-118606.md>), [framework](<https://devfeed.tech/tags/framework.md>), [function](<https://devfeed.tech/tags/function.md>), [pwn2own](<https://devfeed.tech/tags/pwn2own.md>), [remote](<https://devfeed.tech/tags/remote.md>), [remote-code-execution](<https://devfeed.tech/tags/remote-code-execution.md>), [remote-code-execution-vulnerability](<https://devfeed.tech/tags/remote-code-execution-vulnerability.md>), [service](<https://devfeed.tech/tags/service.md>), [vulnerability](<https://devfeed.tech/tags/vulnerability.md>)

### AI overview

The Alpine iLX-F511 framework service exposes a dangerous function that network-adjacent attackers can exploit to execute arbitrary code as root without authentication. ZDI rates the vulnerability CVSS 8.8 and assigns CVE-2026-118606. The advisory says the salient mitigation is to restrict interaction with the product. ZDI reported the issue in February 2026 and publicly released the advisory on October 28, 2026; the excerpt does not describe a vendor fix.

### Source excerpt

This vulnerability allows network-adjacent attackers to execute arbitrary code on affected installations of Alpine iLX-F511 devices. Authentication is not required to exploit this vulnerability. The ZDI has assigned a CVSS rating of 8.8. The following CVEs are assigned: CVE-2026-118606.

## TrustedCode: Evaluating Large Language Models for Developing Trusted Code through Automated Program Verification

DevFeed: [TrustedCode: Evaluating Large Language Models for Developing Trusted Code through Automated Program Verification](<https://devfeed.tech/articles/trustedcode-evaluating-large-language-models-for-developing-trusted-code-through-automated-program-verification-96641.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.30432>)

Author: Muhammad Danial, Muhammad Taimoor Khan

Published: 2026-10-23T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Software Engineering](<https://devfeed.tech/sources/arxiv-software-engineering.md>)

Topics: [Formal verification](<https://devfeed.tech/topics/formal-verification.md>), [Automated reasoning](<https://devfeed.tech/topics/automated-reasoning.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [Chain-of-thought](<https://devfeed.tech/topics/chain-of-thought.md>), [llm-reasoning](<https://devfeed.tech/topics/llm-reasoning.md>), [Code](<https://devfeed.tech/topics/code.md>)

Tags: [automated](<https://devfeed.tech/tags/automated.md>), [chain-of-thought](<https://devfeed.tech/tags/chain-of-thought.md>), [code](<https://devfeed.tech/tags/code.md>), [complexity](<https://devfeed.tech/tags/complexity.md>), [cs-lo](<https://devfeed.tech/tags/cs-lo.md>), [cs-pl](<https://devfeed.tech/tags/cs-pl.md>), [cs-se](<https://devfeed.tech/tags/cs-se.md>), [domain](<https://devfeed.tech/tags/domain.md>), [errors](<https://devfeed.tech/tags/errors.md>), [experiments](<https://devfeed.tech/tags/experiments.md>), [formal](<https://devfeed.tech/tags/formal.md>), [formal-verification](<https://devfeed.tech/tags/formal-verification.md>), [framework](<https://devfeed.tech/tags/framework.md>), [knowledge](<https://devfeed.tech/tags/knowledge.md>), [langgraph](<https://devfeed.tech/tags/langgraph.md>), [large-language-models](<https://devfeed.tech/tags/large-language-models.md>), [prompting](<https://devfeed.tech/tags/prompting.md>), [python](<https://devfeed.tech/tags/python.md>), [software-correctness](<https://devfeed.tech/tags/software-correctness.md>), [specification](<https://devfeed.tech/tags/specification.md>), [trust](<https://devfeed.tech/tags/trust.md>), [type](<https://devfeed.tech/tags/type.md>), [verification](<https://devfeed.tech/tags/verification.md>)

### AI overview

TrustedCode automates formal verification of AI-generated Python: LLMs type-check the code, translate it to WhyML through Why3, then check it against formal specifications with Why3's automated theorem provers. The framework uses chain-of-thought prompting via LangGraph to address challenges in generating specifications and proving programs. In experiments on ten Python programs of varying complexity, reported errors were attributed to domain-knowledge gaps (50%), syntax inaccuracies (30%), and semantic or type issues (20%). These results point to remaining limitations in LLM-guided verification; the abstract does not establish performance across broader benchmarks.

### Source excerpt

arXiv:2610.30432v1 Announce Type: new Abstract: Recent reports, such as "Who is using AI to code?", indicate that approximately 30\% of Python code contributions to GitHub are AI-generated. Given the inherent unreliability of this code, formal verification emerges as a promising approach to establish trust in software correctness. Although leveraging large language models (LLMs) for formal verification has garnered increasing attention, practical adoption remains limited due to the challenges in writing formal specifications and verifying program correctness against them. Existing tools predominantly support verification for languages such as Java and C, with minimal support for Python. We present TrustedCode, an automated framework for verifying AI-generated Python code. TrustedCode uses LLMs to (i) type-check Python code, (ii) translate it into WhyML via the Why3 framework, and (iii) verify it against formal specifications using Why3 back-end automated theorem provers. Our method integrates chain-of-thought prompting through LangGraph to identify and address LLM limitations in specification generation and proving the specified programs. To evaluate TrustedCode, we conducted experiments on ten Python programs of varying complexity. Our analysis reveals that 50\% of the observed errors stem from domain knowledge gaps, 30\% from syntactic inaccuracies, and the remaining 20\% from semantic and type-related issues. These findings underscore the importance of targeted improvements in LLM-guided formal verification workflows for Python. TrustedCode is an open source tool available at: \href{github.com/mtaimoorkhan/TrustedCode}{github.com/mtaimoorkhan/TrustedCode}. The video demonstration is available at: \href{youtu.be/Nr5I2g8ec-s}{youtu.be/Nr5I2g8ec-s}

## Knowledge as Code

DevFeed: [Knowledge as Code](<https://devfeed.tech/articles/knowledge-as-code-93703.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.27693>)

Author: Martin Monperrus

Published: 2026-10-21T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Software Engineering](<https://devfeed.tech/sources/arxiv-software-engineering.md>)

Topics: [Code](<https://devfeed.tech/topics/code.md>), [knowledge-engineering](<https://devfeed.tech/topics/knowledge-engineering.md>)

Tags: [agents](<https://devfeed.tech/tags/agents.md>), [code](<https://devfeed.tech/tags/code.md>), [coding-agents](<https://devfeed.tech/tags/coding-agents.md>), [cs-se](<https://devfeed.tech/tags/cs-se.md>), [encoding](<https://devfeed.tech/tags/encoding.md>), [execution](<https://devfeed.tech/tags/execution.md>), [experiment](<https://devfeed.tech/tags/experiment.md>), [knowledge](<https://devfeed.tech/tags/knowledge.md>), [knowledge-work](<https://devfeed.tech/tags/knowledge-work.md>), [language](<https://devfeed.tech/tags/language.md>), [language-models](<https://devfeed.tech/tags/language-models.md>), [large-language-models](<https://devfeed.tech/tags/large-language-models.md>), [learning](<https://devfeed.tech/tags/learning.md>), [machine](<https://devfeed.tech/tags/machine.md>), [media](<https://devfeed.tech/tags/media.md>), [medium](<https://devfeed.tech/tags/medium.md>), [programming](<https://devfeed.tech/tags/programming.md>), [programming-language](<https://devfeed.tech/tags/programming-language.md>), [properties](<https://devfeed.tech/tags/properties.md>), [show](<https://devfeed.tech/tags/show.md>), [tool](<https://devfeed.tech/tags/tool.md>)

### AI overview

The paper defines "knowledge as code" as knowledge represented in a programming language or machine-parsable data format whose validity can be checked by execution. It distinguishes encoding from learning by execution, in which code serves as both experiment and record. Coding-agent trajectories can make the latter process observable by recording probes, observations and retained code. The authors identify thirteen epistemic properties of knowledge media and argue that code is the first medium to exhibit all of them, making it a knowledge medium that can check what it contains.

### Source excerpt

arXiv:2610.27693v1 Announce Type: new Abstract: For millennia, knowledge has been carried by language, and large language models still put language at the center. Yet the coding agents built on these models mostly write code. This paper is a contribution to epistemology, which asks what kind of knowledge code is and how it is obtained. We define knowledge as code as knowledge written in a programming language or in a machine-parsable data format, whose validity can be checked by executing it. We define thirteen epistemic properties of knowledge media and show that code is the first medium to have all of them. We distinguish two ways in which knowledge enters code, encoding and learning by execution, where code is both the experiment and the record of its result. Agents make learning by execution observable, since their trajectories record every probe, observation and line of code they keep. Code is thus not only a tool of knowledge work but a knowledge medium in its own right, the first that can check what it holds.

## Concluding Debezium's Google Summer of Code 2026 Program

DevFeed: [Concluding Debezium's Google Summer of Code 2026 Program](<https://devfeed.tech/articles/concluding-debezium-s-google-summer-of-code-2026-program-86567.md>)

Original publisher: [Read original article](<https://debezium.io/blog/2026/10/14/gsoc-end/>)

Author: Giovanni Panice

Published: 2026-10-14T00:00:00Z

Content type: news

Language: en

Sources: [Debezium](<https://devfeed.tech/sources/debezium-io.md>)

Topics: [CI/CD](<https://devfeed.tech/topics/cicd.md>), [Code](<https://devfeed.tech/topics/code.md>), [AI Chat](<https://devfeed.tech/topics/ai-chat.md>), [Firestore](<https://devfeed.tech/topics/firestore.md>)

Tags: [code](<https://devfeed.tech/tags/code.md>), [debezium](<https://devfeed.tech/tags/debezium.md>), [developers](<https://devfeed.tech/tags/developers.md>), [google](<https://devfeed.tech/tags/google.md>), [google-summer-of-code](<https://devfeed.tech/tags/google-summer-of-code.md>)

### AI overview

Debezium reports that all its Google Summer of Code 2026 contributors passed their final evaluations. The projects covered CLI tooling, deployment beyond Kubernetes, CDC integration with AI and vector stores, and new Milvus and SQLite connectors; their results are included in Debezium 3.7. The article also encourages contributors to remain involved and previews project posts and a November showcase.

### Source excerpt

The Google Summer of Code 2026 program has officially concluded for Debezium, and we are delighted to announce that all our contributors successfully passed their final evaluations! 🎉 Over the past few months, our contributors have worked closely with their mentors and the wider community to design, implement, test, and document new capabilities across the Debezium ecosystem. Their work was not simply an academic exercise or a collection of experimental prototypes. These contributions address real use cases and are already valuable to the many companies and developers who rely on Debezium. The results are included in Debezium 3.7, which was recently released. Before looking at what comes next, let's take a moment to celebrate the contributors and the projects they delivered. Read more...

## 10X Code ≠ 10X Revenue With Rich Mironov

DevFeed: [10X Code ≠ 10X Revenue With Rich Mironov](<https://devfeed.tech/articles/10x-code-10x-revenue-with-rich-mironov-65749.md>)

Original publisher: [Read original article](<https://dpereira.substack.com/p/10x-code-10x-revenue-with-rich-mironov>)

Author: David Pereira

Published: 2026-10-06T12:30:00Z

Content type: article

Language: en

Sources: [Untrapping Product Teams](<https://devfeed.tech/sources/untrapping-product-teams.md>)

Topics: [Code](<https://devfeed.tech/topics/code.md>)

Tags: [interview](<https://devfeed.tech/tags/interview.md>), [product](<https://devfeed.tech/tags/product.md>)

### AI overview

David Pereira interviews product executive coach Rich Mironov about prioritizing what ships to customers, why more code does not necessarily mean more value, and influencing product decisions.

### Source excerpt

Have you heard about rich mironov ?

## How Rogo ships agent-written code to production in 5 minutes on Vercel

DevFeed: [How Rogo ships agent-written code to production in 5 minutes on Vercel](<https://devfeed.tech/articles/how-rogo-ships-agent-written-code-to-production-in-5-minutes-on-vercel-63817.md>)

Original publisher: [Read original article](<https://vercel.com/blog/how-rogo-ships-agent-written-code-to-production-in-5-minutes-on-vercel>)

Author: Susan Aziz

Published: 2026-10-02T04:00:00Z

Content type: news

Language: en

Sources: [Vercel News](<https://devfeed.tech/sources/vercel-news.md>)

Topics: [Vercel](<https://devfeed.tech/topics/vercel.md>), [AI-assisted coding](<https://devfeed.tech/topics/ai-assisted-coding.md>), [Multi Agent Systems](<https://devfeed.tech/topics/multi-agent-systems.md>), [Code](<https://devfeed.tech/topics/code.md>)

Tags: [000](<https://devfeed.tech/tags/000.md>), [agent](<https://devfeed.tech/tags/agent.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agent](<https://devfeed.tech/tags/ai-agent.md>), [ai-sdk](<https://devfeed.tech/tags/ai-sdk.md>), [analysis](<https://devfeed.tech/tags/analysis.md>), [applications](<https://devfeed.tech/tags/applications.md>), [churn](<https://devfeed.tech/tags/churn.md>), [churn-analysis](<https://devfeed.tech/tags/churn-analysis.md>), [code](<https://devfeed.tech/tags/code.md>), [deal](<https://devfeed.tech/tags/deal.md>), [deployments](<https://devfeed.tech/tags/deployments.md>), [desk](<https://devfeed.tech/tags/desk.md>), [financial](<https://devfeed.tech/tags/financial.md>), [financial-institutions](<https://devfeed.tech/tags/financial-institutions.md>), [institutions](<https://devfeed.tech/tags/institutions.md>), [minutes](<https://devfeed.tech/tags/minutes.md>), [production](<https://devfeed.tech/tags/production.md>), [vercel](<https://devfeed.tech/tags/vercel.md>)

### AI overview

Rogo reports more than 73,000 deployments in a single month, with agent-written code reaching production in five minutes. The company is rearchitecting internal applications on Vercel and uses swarms of agents built with the AI SDK to triage and remediate production incidents without manual intervention. It also reports six production AI agents automating work from churn analysis to deal desk. Rogo builds Felix, which generates decks, financial models, and reports for financial institutions.

### Source excerpt

Rogo on Vercel 73,000+ deployments in a single month Agent-written code shipped to production in 5 minutes 6 production AI agents automating work from churn analysis to deal desk Zero manual triage during production incidents with agent swarms on AI SDK Rogo builds Felix, an AI agent that generates decks, financial models, and reports for a majority of the world's largest financial institutions. The Rogo team is rearchitecting its internal applications on Vercel, where agent-written code ships in 5 minutes. When incidents hit production, swarms of agents on the AI SDK triage and remediate them without manual intervention. As a result, the team now pushes more than 73,000 deployments a month. Read more

## How to Build a Code Graph in TypeScript Using VS Code's Language APIs

DevFeed: [How to Build a Code Graph in TypeScript Using VS Code's Language APIs](<https://devfeed.tech/articles/how-to-build-a-code-graph-in-typescript-using-vs-code-s-language-apis-63485.md>)

Original publisher: [Read original article](<https://www.freecodecamp.org/news/how-to-build-a-code-graph-in-typescript-using-vs-code-language-apis/>)

Author: Otobong Peter

Published: 2026-10-01T11:01:21Z

Content type: tutorial

Language: en

Sources: [Free Code Camp](<https://devfeed.tech/sources/freecodecamp-programming-tutorials-python-javascript-git-more.md>)

Topics: [Visual Studio Code](<https://devfeed.tech/topics/visual-studio-code.md>), [TypeScript](<https://devfeed.tech/topics/typescript.md>), [call hierarchy](<https://devfeed.tech/topics/call-hierarchy.md>), [Code](<https://devfeed.tech/topics/code.md>)

Tags: [code](<https://devfeed.tech/tags/code.md>), [code-graph](<https://devfeed.tech/tags/code-graph.md>), [developer-tools](<https://devfeed.tech/tags/developer-tools.md>), [graphs](<https://devfeed.tech/tags/graphs.md>), [how-to](<https://devfeed.tech/tags/how-to.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [programming-languages](<https://devfeed.tech/tags/programming-languages.md>), [software-development](<https://devfeed.tech/tags/software-development.md>), [typescript](<https://devfeed.tech/tags/typescript.md>), [vs-code](<https://devfeed.tech/tags/vs-code.md>), [vscode](<https://devfeed.tech/tags/vscode.md>)

### AI overview

This tutorial shows how to build a code graph in a VS Code extension using TypeScript and VS Code language APIs. It explains how to discover symbols and call relationships, traverse the graph with breadth-first search, limit concurrency, handle stale language-tooling references, and display the resulting graph in a Webview. It also describes limitations of static call hierarchy analysis and methods for testing the graph builder.

### Source excerpt

Modern codebases are becoming increasingly difficult to navigate. This isn't necessarily because developers are writing more code themselves. It's mostly because coding assistants are generating hundr

## Is your agent writing proper Convex code?

DevFeed: [Is your agent writing proper Convex code?](<https://devfeed.tech/articles/is-your-agent-writing-proper-convex-code-113333.md>)

Original publisher: [Read original article](<https://stack.convex.dev/proper-convex-code-eslint>)

Author: Stack

Published: 2026-10-01T00:06:20Z

Content type: tutorial

Language: en

Sources: [Stack](<https://devfeed.tech/sources/stack.md>)

Topics: [ESLint](<https://devfeed.tech/topics/eslint.md>), [Programming](<https://devfeed.tech/topics/programming.md>), [Code](<https://devfeed.tech/topics/code.md>), [Claude Code](<https://devfeed.tech/topics/claude-code.md>)

Tags: [access-control](<https://devfeed.tech/tags/access-control.md>), [agent](<https://devfeed.tech/tags/agent.md>), [ai](<https://devfeed.tech/tags/ai.md>), [convex](<https://devfeed.tech/tags/convex.md>), [eslint](<https://devfeed.tech/tags/eslint.md>), [patterns](<https://devfeed.tech/tags/patterns.md>)

### AI overview

Convex's ESLint plugin catches Convex-specific mistakes that can pass deployment checks, and its messages often tell coding agents how to fix them. Version 5 can replace `process.env` with typed environment variables when they are declared in `convex.config.ts`, add table names to `db.get` when TypeScript types identify the table, and flag older function styles, missing argument validators, query `.filter` usage, and redundant indexes. Query filters may read increasing amounts of data as tables grow, so the plugin warns that an index may be more appropriate but cannot choose one. Duplicate indexes can add write and storage costs, though different sort orders may justify keeping both. The recommended config covers files in `convex`; Biome cannot load ESLint plugins, while ESLint or Oxlint can run these rules. The author especially recommends enabling the optional `require-access-control` rule: it checks public functions for recognized access-check calls, but its name-based detection is limited and it does not cover HTTP actions or custom functions. Agents must run lint for these checks to help, so the author suggests prompting them to do so in `AGENTS.md` or `CLAUDE.md`.

### Source excerpt

These days my agent writes most of my code, but how do I actually know it's writing good, proper Convex code? One really good way is the Convex ESLint plugin, so here are the rules your agent is going to run into, plus the one I'd really recommend turning on.

## The Second Life of Autoconf for MS-DOS

DevFeed: [The Second Life of Autoconf for MS-DOS](<https://devfeed.tech/articles/the-second-life-of-autoconf-for-ms-dos-61360.md>)

Original publisher: [Read original article](<https://code.dblock.org/2026/09/27/the-second-life-of-autoconf-for-ms-dos.html>)

Author: Daniel Doubrovkine (dblock@dblock.org)

Published: 2026-09-27T00:00:00Z

Content type: article

Language: en

Sources: [Daniel Doubrovkine](<https://devfeed.tech/sources/daniel-doubrovkine.md>)

Topics: [Assembly](<https://devfeed.tech/topics/assembly.md>), [Code](<https://devfeed.tech/topics/code.md>), [llm-reasoning](<https://devfeed.tech/topics/llm-reasoning.md>), [OpenAPI Specification](<https://devfeed.tech/topics/openapi.md>)

Tags: [assembly](<https://devfeed.tech/tags/assembly.md>), [copilot](<https://devfeed.tech/tags/copilot.md>), [programming](<https://devfeed.tech/tags/programming.md>), [vestris](<https://devfeed.tech/tags/vestris.md>), [x86](<https://devfeed.tech/tags/x86.md>)

### AI overview

The author recounts discovering and extending a DOS configuration utility called Autoconf after copying its assembly source from a 1991 French computer magazine. Decades later, scanned archives helped identify the original author, and the listings were transcribed and rebuilt under DOS with Copilot and JWasm.

### Source excerpt

Before GNU Autoconf generated configure scripts, there was a DOS program called Autoconf. It let you keep one CONFIG.SYS and one AUTOEXEC.BAT, then choose among boot configurations by pressing a key. You did not have to wait for a menu. The PC BIOS kept early keystrokes in its keyboard queue, so you could press a configuration letter before AUTOCONF.SYS had even loaded. When the driver finally ran, it found the key and continued immediately. I did not write the original. I copied its x86 assembly source by hand from a French computer magazine. It was the first assembly program I had ever seen and the first piece of code I compiled. I spent the next three or four years extending it into my first commercial product. For a long time I could not remember who had written that first version. I had thrown away the magazine, forgotten the issue, and lost the trail before old print archives became easy to search. Only now, with AI helping sift through a large collection of scanned SVM issues downloaded from Abandonware Magazines and the Internet Archive, was I able to find the seven pages I had copied. The Magazine The original Autoconf appeared in the November 1991 issue of Science & Vie Micro, number 88, on pages 227-233. The article, "La configuration idéale", was written by Julien Pommier and won the magazine's monthly programming contest, including a 2,000-franc prize. The seven pages contained the complete assembly listings for AUTOCONF.ASM and the 11-byte READCONF.COM, installation instructions, sample startup files, and the commands required to build everything with MASM, LINK, and EXE2BIN. I must have spent a week copying the listing without understanding what most of it did, correcting it character by character, and figuring out how to assemble it. I needed multiple boot configurations because I was tired of editing CONFIG.SYS whenever I wanted to play a game. Thirty-five years later, Copilot and I transcribed the listings again and built them under DOS with JWasm.

## Go concurrency distilled

DevFeed: [Go concurrency distilled](<https://devfeed.tech/articles/go-concurrency-distilled-60417.md>)

Original publisher: [Read original article](<https://antonz.org/go-concurrency-distilled/>)

Author: Anton Zhiyanov

Published: 2026-09-26T11:30:00Z

Content type: tutorial

Language: en

Sources: [Anton Zhiyanov](<https://devfeed.tech/sources/anton-zhiyanov.md>)

Topics: [Concurrency](<https://devfeed.tech/topics/concurrency.md>), [Go Language](<https://devfeed.tech/topics/go-language.md>), [Programming](<https://devfeed.tech/topics/programming.md>), [Code](<https://devfeed.tech/topics/code.md>)

Tags: [book](<https://devfeed.tech/tags/book.md>), [code](<https://devfeed.tech/tags/code.md>), [concurrency](<https://devfeed.tech/tags/concurrency.md>), [concurrent-programming](<https://devfeed.tech/tags/concurrent-programming.md>), [go](<https://devfeed.tech/tags/go.md>), [go-concurrency](<https://devfeed.tech/tags/go-concurrency.md>), [interactive](<https://devfeed.tech/tags/interactive.md>), [programming](<https://devfeed.tech/tags/programming.md>)

### AI overview

An interactive mini-book provides a concise refresher on concurrency in Go. It covers goroutines, channels, pipelines, synchronization, data races, testing, scheduling, and diagnostics, with examples that readers can modify and run.

### Source excerpt

Interactive mini-book on concurrent programming in Go.

## Automating DevSecOps Static Analysis with GitHub Actions and Agent Skills

DevFeed: [Automating DevSecOps Static Analysis with GitHub Actions and Agent Skills](<https://devfeed.tech/articles/automating-devsecops-static-analysis-with-github-actions-and-agent-skills-60254.md>)

Original publisher: [Read original article](<https://www.sitepoint.com/automating-devsecops-static-analysis-with-github-actions-and-agent-skills/>)

Author: SitePoint Team

Published: 2026-09-26T05:28:32Z

Content type: tutorial

Language: en

Sources: [SitePoint](<https://devfeed.tech/sources/sitepoint.md>)

Topics: [DevSecOps](<https://devfeed.tech/topics/devsecops.md>), [GitHub Actions](<https://devfeed.tech/topics/github-actions.md>), [Agent Skills](<https://devfeed.tech/topics/agent-skills.md>), [Security](<https://devfeed.tech/topics/security.md>), [GitHub](<https://devfeed.tech/topics/github.md>), [TypeScript](<https://devfeed.tech/topics/typescript.md>), [Vulnerabilities](<https://devfeed.tech/topics/vulnerabilities.md>), [Code](<https://devfeed.tech/topics/code.md>)

Tags: [agent-skills](<https://devfeed.tech/tags/agent-skills.md>), [automated](<https://devfeed.tech/tags/automated.md>), [devops](<https://devfeed.tech/tags/devops.md>), [devsecops](<https://devfeed.tech/tags/devsecops.md>), [github-actions](<https://devfeed.tech/tags/github-actions.md>), [implement](<https://devfeed.tech/tags/implement.md>), [integration](<https://devfeed.tech/tags/integration.md>), [programming](<https://devfeed.tech/tags/programming.md>), [sarif](<https://devfeed.tech/tags/sarif.md>), [security](<https://devfeed.tech/tags/security.md>), [static-analysis](<https://devfeed.tech/tags/static-analysis.md>), [web-security](<https://devfeed.tech/tags/web-security.md>), [workflow](<https://devfeed.tech/tags/workflow.md>)

### AI overview

A tutorial on building an automated DevSecOps static security analysis workflow with GitHub Actions and sandboxed Agent Skills. It describes an AST-based TypeScript vulnerability scanner, SARIF output, pull request annotations, and branch protection enforcement.

### Source excerpt

Implement an automated CI DevSecOps static security analysis workflow in GitHub Actions using sandboxed Agent Skills that outputs standard SARIF reports for native GitHub Code Scanning integration. Continue reading Automating DevSecOps Static Analysis with GitHub Actions and Agent Skills on SitePoint.

## A DOOM Port Demonstrates SQL's Capabilities as a Programming Language

DevFeed: [A DOOM Port Demonstrates SQL's Capabilities as a Programming Language](<https://devfeed.tech/articles/abusing-sql-to-play-doom-60257.md>)

Original publisher: [Read original article](<https://hackaday.com/2026/09/25/abusing-sql-to-play-doom/>)

Author: Bryan Cockfield

Published: 2026-09-26T05:00:05Z

Content type: article

Language: en

Sources: [Hackaday](<https://devfeed.tech/sources/hackaday.md>)

Topics: [SQL](<https://devfeed.tech/topics/sql.md>), [Programming](<https://devfeed.tech/topics/programming.md>), [Code](<https://devfeed.tech/topics/code.md>), [Python](<https://devfeed.tech/topics/python.md>), [GitHub](<https://devfeed.tech/topics/github.md>)

Tags: [code](<https://devfeed.tech/tags/code.md>), [database](<https://devfeed.tech/tags/database.md>), [doom](<https://devfeed.tech/tags/doom.md>), [github](<https://devfeed.tech/tags/github.md>), [programming](<https://devfeed.tech/tags/programming.md>), [python](<https://devfeed.tech/tags/python.md>), [software-hacks](<https://devfeed.tech/tags/software-hacks.md>), [sql](<https://devfeed.tech/tags/sql.md>), [video-game](<https://devfeed.tech/tags/video-game.md>)

### AI overview

A group at CedarDB ported the original DOOM game to SQL, storing its data in tables and implementing rendering with 1,300 lines of SQL across 89 common table expressions. The project also uses Python for keyboard input, timing, and display.

### Source excerpt

Most laypeople who encounter SQL think of it purely as a tool for managing large datasets. While that is certainly what it was designed for, SQL is still a programming ...read more

[Next page](<https://devfeed.tech/topics/code.md?cursor=WyIyMDI2LTA5LTI2VDA1OjAwOjA1KzAwOjAwIiwgIjAyMDU1ZDZjLTcwNjEtNDJjYS04MjY0LTFiYTNhZGMxNDAwYyJd>)