cognitivedaniel commited on
Commit
8c509df
Β·
verified Β·
1 Parent(s): ae10bde

Upload README.md

Browse files
Files changed (1) hide show
  1. README.md +100 -1
README.md CHANGED
@@ -1,3 +1,102 @@
1
  ---
2
- license: mit
 
 
 
 
 
 
 
 
 
 
 
 
3
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
+ license: other
3
+ tags:
4
+ - code
5
+ - agent
6
+ - software-engineering
7
+ - autonomous
8
+ - swe-bench
9
+ - bug-fixing
10
+ - pull-request
11
+ language:
12
+ - en
13
+ - code
14
+ pipeline_tag: text-generation
15
  ---
16
+
17
+ # Daniel β€” Autonomous AI Software Engineer
18
+
19
+ **Daniel** is an autonomous AI software engineer built by [Cognitive AI](https://www.cognitivedaniel.com). It reads your codebase, writes code, runs tests, and opens pull requests β€” without you doing anything manually.
20
+
21
+ πŸ”— **Try it:** [cognitivedaniel.com](https://www.cognitivedaniel.com) | πŸ“Š **Benchmark:** [Daniel vs Devin](https://www.cognitivedaniel.com/devin-alternative)
22
+
23
+ ---
24
+
25
+ ## Benchmark results (v7, October 2026)
26
+
27
+ Daniel was tested head-to-head against Devin on 6 real bugs from popular open-source projects. Both agents received identical prompts. Each PR was graded by the project's own hidden test suite.
28
+
29
+ | Metric | Daniel | Devin |
30
+ |--------|--------|-------|
31
+ | Pass rate | **6/6 (100%)** | 6/6 (100%) |
32
+ | Total cost | **$1.45** | $4.91 |
33
+ | Total time | 12m 33s | 12m 12s |
34
+ | Cost per resolved bug | **$0.24** | $0.82 |
35
+
36
+ **Daniel is 70% cheaper than Devin at the same pass rate.**
37
+
38
+ Dataset: [cognitivedaniel/daniel-benchmark-v1](https://huggingface.co/datasets/cognitivedaniel/daniel-benchmark-v1)
39
+
40
+ ---
41
+
42
+ ## How Daniel works
43
+
44
+ Daniel uses a **4-turn agentic loop**:
45
+
46
+ 1. **Analyze** β€” reads the issue, the failing tests, and the relevant code
47
+ 2. **Implement** β€” makes the minimal code change to fix the bug
48
+ 3. **Test** β€” runs the project's existing test suite to verify the fix
49
+ 4. **PR** β€” opens a pull request with a clear description
50
+
51
+ The loop runs on Anthropic's Claude API. Daniel never opens a PR it hasn't verified against your tests.
52
+
53
+ ---
54
+
55
+ ## Supported languages
56
+
57
+ - Python, JavaScript, TypeScript (primary)
58
+ - Java, Go, Rust (via Multi-SWE-bench tier, in testing)
59
+
60
+ ---
61
+
62
+ ## Integrations
63
+
64
+ - GitHub (repos, issues, PRs)
65
+ - Jira (coming soon)
66
+ - Slack (coming soon)
67
+ - Run locally on any folder with one command
68
+
69
+ ---
70
+
71
+ ## Pricing
72
+
73
+ | Plan | Price | Tasks/month |
74
+ |------|-------|-------------|
75
+ | Starter | $20/mo | 20 |
76
+ | Pro | $100/mo | 200 |
77
+ | Enterprise | $500/mo | Unlimited |
78
+
79
+ ---
80
+
81
+ ## Links
82
+
83
+ - 🌐 [cognitivedaniel.com](https://www.cognitivedaniel.com)
84
+ - πŸ“Š [Benchmark page](https://www.cognitivedaniel.com/devin-alternative)
85
+ - πŸ’» [GitHub](https://github.com/2badagoldman/cognitive-daniel)
86
+ - 🐦 [Twitter / X](https://x.com/CognitiveDaniel)
87
+
88
+ ---
89
+
90
+ ## Citation
91
+
92
+ If you use Daniel's benchmark results in your research:
93
+
94
+ ```bibtex
95
+ @misc{cognitive-ai-daniel-2026,
96
+ title={Daniel: Autonomous AI Software Engineer β€” Public Benchmark v1},
97
+ author={Cognitive AI},
98
+ year={2026},
99
+ url={https://www.cognitivedaniel.com/devin-alternative},
100
+ note={Dataset: huggingface.co/datasets/cognitivedaniel/daniel-benchmark-v1}
101
+ }
102
+ ```