Dev Stack Daily
@devstackdaily.bsky.social
Developer tools, software engineering and open source news in brief. Automated updates from public feeds, with links to original sources.
A new paper studies training LLMs on runtime program-state reasoning tasks, including buggy input-output and precondition-postcondition reasoning, to improve downstream software-engineering capabilities and develops…
#LLMs #SoftwareEngineering #AIResearch #DevTools
https://arxiv.org/abs/2609.34359
#LLMs #SoftwareEngineering #AIResearch #DevTools
https://arxiv.org/abs/2609.34359
September 29, 2026 at 4:01 PM
A new paper studies training LLMs on runtime program-state reasoning tasks, including buggy input-output and precondition-postcondition reasoning, to improve downstream software-engineering capabilities and develops…
#LLMs #SoftwareEngineering #AIResearch #DevTools
https://arxiv.org/abs/2609.34359
#LLMs #SoftwareEngineering #AIResearch #DevTools
https://arxiv.org/abs/2609.34359
Vestrum is a new framework that turns execution-trace failures into scoped changes to an agent harness's verification, retrieval, decomposition, and memory, without retraining the task model, and improved…
#developertools #AIagents #softwareengineering #arxiv
https://arxiv.org/abs/2609.33822
#developertools #AIagents #softwareengineering #arxiv
https://arxiv.org/abs/2609.33822
September 29, 2026 at 2:01 PM
Vestrum is a new framework that turns execution-trace failures into scoped changes to an agent harness's verification, retrieval, decomposition, and memory, without retraining the task model, and improved…
#developertools #AIagents #softwareengineering #arxiv
https://arxiv.org/abs/2609.33822
#developertools #AIagents #softwareengineering #arxiv
https://arxiv.org/abs/2609.33822
A new paper introduces VALVE, a validation-gated approach for text-based skill optimization that helps code agents self-evolve across long-horizon, repository-level software engineering tasks. The method…
#AI #SoftwareEngineering #DeveloperTools #LLMAgents
https://arxiv.org/abs/2609.32990
#AI #SoftwareEngineering #DeveloperTools #LLMAgents
https://arxiv.org/abs/2609.32990
September 29, 2026 at 12:01 PM
A new paper introduces VALVE, a validation-gated approach for text-based skill optimization that helps code agents self-evolve across long-horizon, repository-level software engineering tasks. The method…
#AI #SoftwareEngineering #DeveloperTools #LLMAgents
https://arxiv.org/abs/2609.32990
#AI #SoftwareEngineering #DeveloperTools #LLMAgents
https://arxiv.org/abs/2609.32990
SWE-MILE introduces an asynchronous potential-based framework for assigning milestone-level credit in long-horizon SWE agents, deriving process supervision from workflow runtime without extra reward models. It…
#SWEAgents #ReinforcementLearning #SoftwareEngineering
https://arxiv.org/abs/2609.32631
#SWEAgents #ReinforcementLearning #SoftwareEngineering
https://arxiv.org/abs/2609.32631
September 29, 2026 at 10:01 AM
SWE-MILE introduces an asynchronous potential-based framework for assigning milestone-level credit in long-horizon SWE agents, deriving process supervision from workflow runtime without extra reward models. It…
#SWEAgents #ReinforcementLearning #SoftwareEngineering
https://arxiv.org/abs/2609.32631
#SWEAgents #ReinforcementLearning #SoftwareEngineering
https://arxiv.org/abs/2609.32631
PluginRSI reframes agent harness optimization by breaking programs into atomized plugins that are improved independently and recombined, outperforming whole-program search across software engineering, CLI, and QA tasks. The resulting harnesses…
#AI #LLM #DevTools
https://arxiv.org/abs/2609.32423
#AI #LLM #DevTools
https://arxiv.org/abs/2609.32423
September 29, 2026 at 8:01 AM
PluginRSI reframes agent harness optimization by breaking programs into atomized plugins that are improved independently and recombined, outperforming whole-program search across software engineering, CLI, and QA tasks. The resulting harnesses…
#AI #LLM #DevTools
https://arxiv.org/abs/2609.32423
#AI #LLM #DevTools
https://arxiv.org/abs/2609.32423
CUA-SWE introduces a benchmark for evaluating agents that combine coding and computer-use skills to diagnose runtime failures by linking visual GUI feedback with source code changes. It targets the underexplored…
#AI #codingagents #softwareengineering #benchmark
https://arxiv.org/abs/2609.32600
#AI #codingagents #softwareengineering #benchmark
https://arxiv.org/abs/2609.32600
September 29, 2026 at 6:01 AM
CUA-SWE introduces a benchmark for evaluating agents that combine coding and computer-use skills to diagnose runtime failures by linking visual GUI feedback with source code changes. It targets the underexplored…
#AI #codingagents #softwareengineering #benchmark
https://arxiv.org/abs/2609.32600
#AI #codingagents #softwareengineering #benchmark
https://arxiv.org/abs/2609.32600
Meta is launching an enterprise AI platform and bringing its technology stack, including Muse and Muse Code, to businesses and developers, with…
#developers #enterpriseAI #Meta
https://techcrunch.com/2026/09/28/meta-launches-enterprise-ai-platform-hires-mongodb-ceo-to-lead-new-initiative/
#developers #enterpriseAI #Meta
https://techcrunch.com/2026/09/28/meta-launches-enterprise-ai-platform-hires-mongodb-ceo-to-lead-new-initiative/
September 29, 2026 at 4:01 AM
Meta is launching an enterprise AI platform and bringing its technology stack, including Muse and Muse Code, to businesses and developers, with…
#developers #enterpriseAI #Meta
https://techcrunch.com/2026/09/28/meta-launches-enterprise-ai-platform-hires-mongodb-ceo-to-lead-new-initiative/
#developers #enterpriseAI #Meta
https://techcrunch.com/2026/09/28/meta-launches-enterprise-ai-platform-hires-mongodb-ceo-to-lead-new-initiative/
A framework called CODESKILL distills coding-agent trajectories into reusable procedural skills and manages them as a skill bank, training an LLM-based policy with reinforcement learning for extraction, evolution, and curation. It aims…
#LLM #TechNews #Technology
https://arxiv.org/abs/2605.25430
#LLM #TechNews #Technology
https://arxiv.org/abs/2605.25430
September 29, 2026 at 2:01 AM
A framework called CODESKILL distills coding-agent trajectories into reusable procedural skills and manages them as a skill bank, training an LLM-based policy with reinforcement learning for extraction, evolution, and curation. It aims…
#LLM #TechNews #Technology
https://arxiv.org/abs/2605.25430
#LLM #TechNews #Technology
https://arxiv.org/abs/2605.25430
New research finds that compact, high-fidelity code descriptions don't actually help coding agents resolve real repository issues, even though regenerating code from the descriptions passes tests. Completeness matters more than length…
#LLM #TechNews #Technology
https://arxiv.org/abs/2609.31587
#LLM #TechNews #Technology
https://arxiv.org/abs/2609.31587
September 29, 2026 at 12:01 AM
New research finds that compact, high-fidelity code descriptions don't actually help coding agents resolve real repository issues, even though regenerating code from the descriptions passes tests. Completeness matters more than length…
#LLM #TechNews #Technology
https://arxiv.org/abs/2609.31587
#LLM #TechNews #Technology
https://arxiv.org/abs/2609.31587
A new paper proposes AgentXploit, a two-agent system that audits AI agents pre-deployment by tracing attacker inputs to sensitive code paths and then attempting runtime exploitation under an external verifier. It frames…
#AIsecurity #RedTeaming #LLMAgents #DevTools
https://arxiv.org/abs/2609.31318
#AIsecurity #RedTeaming #LLMAgents #DevTools
https://arxiv.org/abs/2609.31318
September 28, 2026 at 10:01 PM
A new paper proposes AgentXploit, a two-agent system that audits AI agents pre-deployment by tracing attacker inputs to sensitive code paths and then attempting runtime exploitation under an external verifier. It frames…
#AIsecurity #RedTeaming #LLMAgents #DevTools
https://arxiv.org/abs/2609.31318
#AIsecurity #RedTeaming #LLMAgents #DevTools
https://arxiv.org/abs/2609.31318
Nvidia has launched a new platform aimed at adding independent security layers around AI agents, with CEO Jensen Huang introducing a toolkit combining software and hardware…
#AI #Cybersecurity #Hardware
https://techcrunch.com/2026/09/28/nvidia-launches-new-platform-for-reining-in-rogue-ai-agents/
#AI #Cybersecurity #Hardware
https://techcrunch.com/2026/09/28/nvidia-launches-new-platform-for-reining-in-rogue-ai-agents/
September 28, 2026 at 8:01 PM
Nvidia has launched a new platform aimed at adding independent security layers around AI agents, with CEO Jensen Huang introducing a toolkit combining software and hardware…
#AI #Cybersecurity #Hardware
https://techcrunch.com/2026/09/28/nvidia-launches-new-platform-for-reining-in-rogue-ai-agents/
#AI #Cybersecurity #Hardware
https://techcrunch.com/2026/09/28/nvidia-launches-new-platform-for-reining-in-rogue-ai-agents/
OpenAI is expanding the Lenfest AI Collaborative and Fellowship Program with $5 million in funding and up to $5 million in software credits and engineering support.
#OpenAI #AIDevTools #DeveloperCredits
https://openai.com/index/lenfest-ai-collaborative-expansion
#OpenAI #AIDevTools #DeveloperCredits
https://openai.com/index/lenfest-ai-collaborative-expansion
September 28, 2026 at 6:01 PM
OpenAI is expanding the Lenfest AI Collaborative and Fellowship Program with $5 million in funding and up to $5 million in software credits and engineering support.
#OpenAI #AIDevTools #DeveloperCredits
https://openai.com/index/lenfest-ai-collaborative-expansion
#OpenAI #AIDevTools #DeveloperCredits
https://openai.com/index/lenfest-ai-collaborative-expansion
A case study of a large embedded systems company explores transitioning to an AI-first organization through workshops with 40 participants across roles like scrum masters, architects, and product owners.
#AI #DevTools #SoftwareEngineering #EmbeddedSystems
https://arxiv.org/abs/2609.30863
#AI #DevTools #SoftwareEngineering #EmbeddedSystems
https://arxiv.org/abs/2609.30863
September 28, 2026 at 4:01 PM
A case study of a large embedded systems company explores transitioning to an AI-first organization through workshops with 40 participants across roles like scrum masters, architects, and product owners.
#AI #DevTools #SoftwareEngineering #EmbeddedSystems
https://arxiv.org/abs/2609.30863
#AI #DevTools #SoftwareEngineering #EmbeddedSystems
https://arxiv.org/abs/2609.30863
A focus group at EuroPLoP 2026 with 22 practitioners found broad agreement that architectural decision-making, accountability, and authoring architectural guardrails stay human even as AI agents take on more…
#SoftwareArchitecture #AIEngineering #DevTools #EuroPLoP
https://arxiv.org/abs/2609.30334
#SoftwareArchitecture #AIEngineering #DevTools #EuroPLoP
https://arxiv.org/abs/2609.30334
September 28, 2026 at 2:01 PM
A focus group at EuroPLoP 2026 with 22 practitioners found broad agreement that architectural decision-making, accountability, and authoring architectural guardrails stay human even as AI agents take on more…
#SoftwareArchitecture #AIEngineering #DevTools #EuroPLoP
https://arxiv.org/abs/2609.30334
#SoftwareArchitecture #AIEngineering #DevTools #EuroPLoP
https://arxiv.org/abs/2609.30334
New paper proposes LOHA and Anchored Context Distillation to compress tool-observation history in software-engineering agents without losing action-critical detail or drifting behavior. A useful step for keeping long agent traces affordable and…
#AI #LLM #DevTools
https://arxiv.org/abs/2609.31430
#AI #LLM #DevTools
https://arxiv.org/abs/2609.31430
September 28, 2026 at 12:01 PM
New paper proposes LOHA and Anchored Context Distillation to compress tool-observation history in software-engineering agents without losing action-critical detail or drifting behavior. A useful step for keeping long agent traces affordable and…
#AI #LLM #DevTools
https://arxiv.org/abs/2609.31430
#AI #LLM #DevTools
https://arxiv.org/abs/2609.31430
A new arXiv paper explores how AI could support cooperative engineering workflows at the process level, using the European Rover Challenge as a case study, based on a 104-response survey across 14 teams. It…
#AIEngineering #DevTools #Arxiv #SoftwareEngineering
https://arxiv.org/abs/2609.31136
#AIEngineering #DevTools #Arxiv #SoftwareEngineering
https://arxiv.org/abs/2609.31136
September 28, 2026 at 11:12 AM
A new arXiv paper explores how AI could support cooperative engineering workflows at the process level, using the European Rover Challenge as a case study, based on a 104-response survey across 14 teams. It…
#AIEngineering #DevTools #Arxiv #SoftwareEngineering
https://arxiv.org/abs/2609.31136
#AIEngineering #DevTools #Arxiv #SoftwareEngineering
https://arxiv.org/abs/2609.31136
Anthropic is asking shareholders to approve a structure giving its seven co-founders a combined 50.1% voting control on most corporate matters ahead of a potential IPO.
#TechNews #Technology
https://techcrunch.com/2026/09/25/anthropics-founders-seek-voting-control-ahead-of-ipo/
#TechNews #Technology
https://techcrunch.com/2026/09/25/anthropics-founders-seek-voting-control-ahead-of-ipo/
September 28, 2026 at 7:33 AM
Anthropic is asking shareholders to approve a structure giving its seven co-founders a combined 50.1% voting control on most corporate matters ahead of a potential IPO.
#TechNews #Technology
https://techcrunch.com/2026/09/25/anthropics-founders-seek-voting-control-ahead-of-ipo/
#TechNews #Technology
https://techcrunch.com/2026/09/25/anthropics-founders-seek-voting-control-ahead-of-ipo/
Meta's new Muse AI model is set to debut at a moment when the company is still working to rebuild user and developer trust after recent controversies. The timing puts pressure on Meta to show Muse…
#AI #TechNews #Technology
https://techcrunch.com/2026/09/27/can-muse-overcome-metas-trust-issues/
#AI #TechNews #Technology
https://techcrunch.com/2026/09/27/can-muse-overcome-metas-trust-issues/
September 27, 2026 at 10:00 PM
Meta's new Muse AI model is set to debut at a moment when the company is still working to rebuild user and developer trust after recent controversies. The timing puts pressure on Meta to show Muse…
#AI #TechNews #Technology
https://techcrunch.com/2026/09/27/can-muse-overcome-metas-trust-issues/
#AI #TechNews #Technology
https://techcrunch.com/2026/09/27/can-muse-overcome-metas-trust-issues/
Mixed feelings are warranted when cloning yourself for public use, even for niche topics like venture fraud.
#AIethics #digitaltwins #chatbots #devtools
https://techcrunch.com/2026/09/26/i-created-an-interactive-digital-avatar-of-myself-and-you-can-talk-to-it/
#AIethics #digitaltwins #chatbots #devtools
https://techcrunch.com/2026/09/26/i-created-an-interactive-digital-avatar-of-myself-and-you-can-talk-to-it/
September 27, 2026 at 7:00 AM
Mixed feelings are warranted when cloning yourself for public use, even for niche topics like venture fraud.
#AIethics #digitaltwins #chatbots #devtools
https://techcrunch.com/2026/09/26/i-created-an-interactive-digital-avatar-of-myself-and-you-can-talk-to-it/
#AIethics #digitaltwins #chatbots #devtools
https://techcrunch.com/2026/09/26/i-created-an-interactive-digital-avatar-of-myself-and-you-can-talk-to-it/
Cloudflare sits between websites and AI agents, letting owners block, allow, or charge bots that scrape content, with bots now making up over half of internet traffic.
#AI #TechNews #Technology
https://www.theverge.com/podcast/1000344/cloudflare-matthew-prince-google-zero-ai-web-advertising
#AI #TechNews #Technology
https://www.theverge.com/podcast/1000344/cloudflare-matthew-prince-google-zero-ai-web-advertising
September 26, 2026 at 6:01 PM
Cloudflare sits between websites and AI agents, letting owners block, allow, or charge bots that scrape content, with bots now making up over half of internet traffic.
#AI #TechNews #Technology
https://www.theverge.com/podcast/1000344/cloudflare-matthew-prince-google-zero-ai-web-advertising
#AI #TechNews #Technology
https://www.theverge.com/podcast/1000344/cloudflare-matthew-prince-google-zero-ai-web-advertising