Releases·Tool Release
Show HN: Nightshift – Rust CLI to Orchestrate GitHub Issue Resolution with DAGs
A developer released Nightshift, an agent-agnostic Rust CLI that orchestrates completion of GitHub Issues using DAGs. The tool was built after the author found OpenAI and Anthropic's `/goal` feature inadequate for complex multi-issue workflows due to context rot. It gives practitioners a way to manage long-horizon agent tasks without relying on the vendors' context management.
4 engagement1 sourceThu, Sep 10, 2026, 01:01 PM
AI Overview
- Nightshift, an agent-agnostic Rust CLI, orchestrates GitHub issue resolution by decomposing PRDs into child issues with kanban-style dependencies and launching fresh agent sessions per issue
- The tool targets context rot in /goal-style long-horizon runs: author says LLM reasoning degrades beyond ~150k tokens and compaction causes amnesia, hallucinated decisions, and unreliable context
- Each session gets only PRD plus one issue's context piped via stdin; state lives entirely in filesystem and git, yielding deterministic scheduling and failure isolation
- Nightshift supports claude-code, codex, cursor, antigravity, pi, opencode, and github-copilot, with more agents planned
Entities
Related
News·Marketsfeatured
Leaked figures show OpenAI's revenue run rate nears $70B as enterprise sales double
Bloomberg reports OpenAI is targeting $30 billion in new funding at a $1.4 trillion valuation. Axios separately reports OpenAI's annualized revenue run rate is approaching $70 billion, up more than 70% since the start of Q3, with enterprise revenue more than doubling since July. The figures land as OpenAI and Anthropic prepare for potential IPOs.
Bloomberg reports OpenAI is targeting $30 billion in new funding at a $1.4 trillion valuation. Axios separately reports OpenAI's annualized revenue run rate is approaching $70 billion, up more than 70% since the start of Q3, with enterprise revenue more than doubling since July. The figures land as OpenAI and Anthropic prepare for potential IPOs.
Updated Tue, Sep 29, 2026, 07:52 PM — Bloomberg adds OpenAI is targeting $30B in new funding at a $1.4T valuation alongside the $70B revenue figures.
020337 engagement4 sourcesreddit, hackernews, rssTue, Sep 29, 2026, 02:33 PM
Developments·Paper
Chroma tests 18 models, finds context rot degrades agents before window fills
Chroma tested 18 models and found every one became less reliable as input grew, even on simple retrieval, with degradation starting well before the context window is full. The finding challenges the reflex to blame the model or swap frameworks for agent failures on long tasks, pointing instead to context rot as the culprit.
Chroma tested 18 models and found every one became less reliable as input grew, even on simple retrieval, with degradation starting well before the context window is full. The finding challenges the reflex to blame the model or swap frameworks for agent failures on long tasks, pointing instead to context rot as the culprit.
12 engagement1 sourceredditWed, Aug 12, 2026, 01:46 PM
Social·Community
Developer building Codex memory layer says single CRITICAL task burned half a 5-hour allowance on repo rediscovery
A developer building an agent workflow around OpenAI Codex for a Laravel/React project reports that orchestration works but the context/memory side does not, and asks for alternatives to repeatedly re-reading the repo. The post highlights a practical gap in persistent memory tooling for coding-agent pipelines.
A developer building an agent workflow around OpenAI Codex for a Laravel/React project reports that orchestration works but the context/memory side does not, and asks for alternatives to repeatedly re-reading the repo. The post highlights a practical gap in persistent memory tooling for coding-agent pipelines.
18 engagement1 sourceredditWed, Sep 9, 2026, 09:56 PM
Social·Community
Developer moves Codex/Claude run state to disk to avoid unreviewable marathon chats
A developer describes a workflow for long coding-agent runs where the conversation becomes too large to review, so they persist the run's state to disk. This allows them to answer questions about progress, decisions, and blockers without digging through a huge chat, and to resume cleanly after a session dies.
A developer describes a workflow for long coding-agent runs where the conversation becomes too large to review, so they persist the run's state to disk. This allows them to answer questions about progress, decisions, and blockers without digging through a huge chat, and to resume cleanly after a session dies.
15 engagement1 sourceredditWed, Aug 26, 2026, 02:51 PM