← The Independent 100: a study

Charles Goddard

#25 in The Frontier Outside, the curator’s rank and section @chargoddard Relevance to our work: 4 of 5

Bracketed tags are evidence labels: [Verified] means we read the page, file or API response ourselves, [Reported] means a source says so and we did not check the underlying thing, and [Inference] is our own reasoning. To pin is to freeze a dependency, oracle, or assessment at an exact commit or version, so later changes cannot silently alter what was checked.

Credit

The Independent 100 list was curated by dan (@irl_danB) and published on 2026-09-24 at independent.prose.md (announcement). The ranks and section names are his, and number 100 is his open wild-card slot. This study is ours and is not endorsed by him or by anyone on the list.

What they publicly do

Chief of Frontier Research at Arcee AI and author of MergeKit [Reported: X profile x.com/chargoddard]; top MergeKit committer (251 commits) [Verified via GitHub contributors API]. Recently open-sourced Arcee's internal distillation tool in DistillKit and wrote on converting AFM-4.5B to hybrid Kimi Delta Attention [Verified: x.com/chargoddard/status/2000656978046279833].

Public work

  • MergeKit repo LGPL-3.0 last activity 2026-09-12 · 7,371 GitHub stars on 2026-09-24

    Toolkit for merging pretrained LLM weights (SLERP, TIES, DARE etc.) [Verified]

  • DistillKit repo Apache-2.0 last activity 2026-05-12 · 1,069 GitHub stars on 2026-09-24

    Online and offline knowledge distillation with a logprob compression scheme [Verified via his post and repo]

  • Distilling Kimi Delta Attention into AFM-4.5B blog no license stated last activity 2025-12-15

    Post-hoc architecture conversion via distillation [Verified link from his post; post body not read]

  • bitnet repo MIT last activity 2024-04-30 · 24 GitHub stars on 2026-09-24

    Modeling code for a BitNet b1.58 Llama-style model [Verified]

  • pptx-api repo no license file last activity 2025-04-20 · 3 GitHub stars on 2026-09-24

    Generate PPTX from a JSON schema, for LLMs [Verified]

Licenses of GitHub repositories are from the authenticated GitHub license API on 2026-09-25, one row per repository in evidence/licenses.jsonl (“no license file” means the API found none; “license file present, not recognised by GitHub” means GitHub could not classify it). Other licenses are as the source states. Check the license file before reusing anything.

Relevance to our work

4 of 5 lenses: training, local-models

MergeKit and DistillKit are the reference open tools for merging and distilling small models, which is how custom local models get made; for teams that do little training, this is study-first rather than adopt-now.

Scored 0 to 5 against the study’s relevance rubric: 5 means public, openly licensed work squarely in our lanes that we could study or adopt now, and 1 means little connection to our work. The score is one reviewer’s judgement of how much this public work bears on our own work. It is not a ranking of people and says nothing about the person or the quality of their work.

What we would study next

  1. Read the KDA-into-AFM distillation post and DistillKit's logprob compression (arcee.ai/blog/distilling-kimi-delta-attention-into-afm-4-5b-and-the-tool-we-used-to-do-it). Shows how a small team distills a new architecture cheaply; relevant to anyone building small specialist models Effort: small.
  2. Try a MergeKit merge of two MLX-convertible models (github.com/arcee-ai/mergekit). Cheapest way to test whether merged local models beat single ones on a given eval set Effort: medium.

Before reusing anything: MergeKit was relicensed to BSL on 2025-02-05 (commit 'MergeKit v0.1 ... BSL license') and returned to LGPL-3.0 on 2025-10-31 [Verified via LICENSE commit history; also x.com/chargoddard/status/1984321399310172171]. Releases from that window are under BSL, not OSI; pin to a post-2025-10-31 version. LGPL obligations apply if modified and redistributed. cg123/rathe is NOASSERTION; pptx-api has no license.

Sources

Overall confidence in this record: High. Sources were read on 2026-09-24 and every link was checked on 2026-09-25. The full record is line 25 of people.jsonl.

Limits

Not verified: Did not read the Arcee blog post body or verify Arcee's current model releases and their licenses. goddard.blog has no posts yet.

This record comes from a first-pass survey of public material by an AI agent session on 2026-09-24, edited for publication on 2026-09-25. Nothing was installed or run for it. Roles, stars and licenses change; see the study’s limits.

Request a correction

Is something about this record wrong or out of date, or would you rather not be covered? Open a correction (a GitHub issue form with this page’s address filled in). A request to remove a record is honoured without argument; accepted corrections are recorded in the changelog with the date.

Help fix it

Each opens a GitHub issue form with this page’s address filled in. You need a GitHub account.

Made by Joshua Nowak (ZestStream) with AI coding agents. Not affiliated with or funded by Jeffrey Emanuel; Joshua uses his public tools and is a paying subscriber to jeffreys-skills.md.

Source on GitHub MIT licensed