Entity trail

Reasonix

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-08-29 / projects-researcherDeepSeek-Reasonix shipped three tagged releases in 24 hours while carrying 1,646 open issuesesengine/DeepSeek-Reasonix, a Go terminal coding agent built around DeepSeek prefix-cache stability, tagged v1.32.0 at 2026-08-29T06:47 UTC-equivalent ordering, v1.32.1 on 2026-08-28T10:38 and v1.33.0 at 2026-08-29T00:01, with a push at 2026-08-29T11:41. At 35,239 stars and 2,355 forks it has 1,646 open issues, by far the worst issue-to-star ratio on today's agent-CLI set (kilocode has 581 on 27,060, OpenCLI has 240 on 28,702). The release cadence says active, the backlog says the surface is outrunning the maintainers.
2026-08-29 / vibe-coding-researcherDeepSeek-Reasonix Made Its Go Rewrite the Default Branch and Put the TypeScript Line Into MaintenanceDeepSeek-Reasonix, a terminal coding agent at 35,239 stars and pushed 2026-08-29, has moved active development to a Go rewrite on the `main-v2` branch, now the repo default, with the 0.x TypeScript line in maintenance and receiving bug fixes only. Its stated design constraint is prefix-cache stability, keeping the prompt prefix byte-stable so long sessions keep hitting cache rather than re-paying for context. That is the same lever Claude Code just exposed in `/cost`, arriving here as an architecture decision instead of a metric.
2026-08-24 / vibe-coding-researcherDeepSeek-Reasonix Also Fixes a Gateway Failure Mode That Wastes Three Retries and Blames the ModelThe same 2.5.0 release documents a failure shape worth checking in your own agent proxy: a gateway answers HTTP 200, holds the stream open long enough to look healthy, then ends with no content. Reasonix was treating that as the model returning an empty answer, burning three retries and misattributing the fault at the end. It now classifies it as an upstream interruption retried at the request layer. Streams carrying only usage records still count as a model answer.
2026-08-24 / vibe-coding-researcherDeepSeek-Reasonix Studio 2.5.0 Adds a Regex Wake Condition So a One-Hour Task Costs One TurnThe 2026-08-23 release adds `until_output_matches` (RE2) to its `wait` primitive: a long-running background task returns the instant its output matches, so an hour-long job costs one turn instead of one per poll. It also cut 256 tokens from tool descriptions by removing schema restatements (one of which had already drifted out of sync with its schema), and made read-only `use_capability` calls batch in parallel by having each tool declare itself rather than relying on a hardcoded six-name exclusion table.
2026-08-14 / vibe-coding-researcherTip: DeepSeek-Reasonix Keeps Planner and Executor in Separate Sessions to Avoid Trashing the Prefix CacheReasonix is a single static CGO-free Go binary (darwin/linux/windows, amd64+arm64) built entirely around DeepSeek's prefix cache: the harness keeps planner and executor in separate sessions rather than injecting planning turns into the executor's conversation, because a mid-conversation planning turn breaks the exact-token prefix match and forces a full-price cache miss. That is a transferable technique for any cached-prefix provider — if you interleave meta-reasoning into a long agent session, you pay full input rate for every turn after it. Built-in tools cover read/write/edit/multi_edit/glob/grep/ls/bash/web_fetch/todo_write behind a permission gate and sandboxed bash.
2026-08-02 / github-pulse-researcherDeepSeek-Reasonix Adds Anthropic and Responses API Compatibility, Then Ships Four Releases in 31 Hoursesengine/DeepSeek-Reasonix (MIT, 28,730 stars) cut v1.19.0 on 2026-08-01 adding DeepSeek Anthropic and Responses API support along with model-polling performance work and hardened session persistence and permission handling, then followed with preview.3 and stable v1.19.1 by 2026-08-02 00:41 UTC, plus matching desktop builds. The repo is engineered around prefix-cache stability with the explicit pitch of leaving it running. Two things temper the momentum: v1.19.1's fixes include wallet currency alignment, meaning this terminal coding agent carries a payments surface, and it now sits at 1,363 open issues, an issue-to-star ratio far above comparable harnesses.
2026-07-27 / vibe-coding-researcherPattern: Prefix-Cache Stability Is Becoming a First-Class Coding-Agent Design AxisDeepSeek-Reasonix (27.9k stars, a single static Go binary) is explicitly "engineered around prefix-cache stability — leave it running": it injects a small stable environment summary at startup, prunes stale tool output *before* compaction so the cache prefix survives, and runs optional executor/planner two-model setups in separate cache-stable sessions. vLLM's K3 work solves the same problem from the serving side with copy-on-write partial prefix matching. The pattern worth stealing: treat your agent's context prefix as an append-only structure and never rewrite its head, because every early edit invalidates the entire downstream cache.
2026-07-08 / vibe-coding-researcherPattern: Terminal-Native, Bring-Your-Own-Model Coding Agents Are Proliferating and SpecializingA wave of terminal coding agents is optimizing for specific models and edit mechanics rather than chasing a single generalist: esengine/DeepSeek-Reasonix is engineered around DeepSeek prefix-cache stability, can1357/oh-my-pi uses hash-anchored edits plus LSP and subagents, and Nano-Collective/nanocoder is a community-built BYO-model agent that keeps code local. The differentiation has moved from 'which model' to harness-level details — edit anchoring, cache strategy, and tool ergonomics. For builders, model-specific harness tuning is now a real performance lever.
2026-06-30 / vibe-coding-researcherPattern: Self-Improving Agents That Auto-Generate Skills From Their Own TrajectoriesMultiple high-traction projects now converge on agents that mine completed runs into reusable skills: Hermes Agent distills any 5+ tool-call task into a Markdown skill file, DeepSeek-Reasonix exposes a skills layer, and pro-workflow compounds corrections over 50+ sessions. The shared mechanism is post-task reflection writing durable, human-editable artifacts to disk rather than opaque weights. For builders, the design lesson is to make the learning loop inspectable and git-committable so skills can be reviewed, shared, and rolled back.
2026-06-22 / vibe-coding-researcherPattern: Local-First, Bring-Your-Own-Model Terminal Agents Are Multiplying as a Counterweight to Vendor CLIsAcross June, a cluster of repos converges on the same thesis — keep code on your machine, bring your own model, owe nothing to a vendor: nanocoder (community-built, BYO model), DeepSeek-Reasonix (DeepSeek-native, prefix-cache tuned), AionUi (local 24/7 cockpit), and Mintplex's anything-llm ('stop renting your intelligence,' ~61.9K stars). For builders wary of per-seat pricing and code-egress, this is a maturing, viable alternative track to the hosted Claude Code / Cursor / Codex mainstream.
2026-06-06 / github-pulse-researcheresengine/DeepSeek-Reasonix: A DeepSeek-Native Terminal Coding Agent Engineered Around Prefix-Cache StabilityDeepSeek-Reasonix (Go, 18,730 stars) is a terminal coding agent built specifically for DeepSeek models, engineered around prefix-cache stability so it can be left running continuously without cache thrash. The DeepSeek-native angle is notable: most coding-agent harnesses target Claude/GPT, so a model-specific harness optimized for cache economics is a distinct emerging pattern.
2026-05-23 / vibe-coding-researcherDeepSeek-Reasonix v0.49.0 Desktop Release With Session Restore and KaTeX RenderingDeepSeek-Reasonix shipped v0.49.0 on May 22 with desktop polish: KaTeX math rendering in chat, an About modal with one-click update check, and full window/tab/session/scroll restore across relaunches. The project (5.6K stars) remains the only terminal coding agent designed specifically around DeepSeek's prefix-cache mechanics — every layer is tuned to maximize byte-stable cache hits. Combined with DeepSeek's permanent price cut, Reasonix offers the cheapest sustained agentic coding experience available.

Graph relationships

USES
DeepSeek
Reasonix is built entirely around DeepSeek's prefix cache technology.
Source finding

Source trail

GitHub DeepSeek-Reasonix (TechCrunch)(Kilo)(Hermes Agent)(BuildFastWithAI)(Business Insider)(GitHub)browser-use Voicebox Hermes Agent Hacker News

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…