Advanced AI & Agentic Systems
PASSGPA3.07AVG78.2RETAKES
A1 GPA 3.07 · avg 78.2 · B+ B C+ C+ A- B A- B+
Holt detail live2m ago
live/campaign sample 5s ago · Aug 8, 2026 · 3:23:30 AM EDT
transcript/history sample 2m ago · Aug 8, 2026 · 3:21:32 AM EDT
Reason-eval re-grade
PURPOSEItem-by-item re-grade of Holt's originally missed questions, retained as part of the final graduation record.
CURRENT STATUS COMPLETE · re-grade folded into final transcript · Jul 22, 2026 · 1:50:22 AM EDT
127/127 items scored
Re-grade complete · folded into final transcript Jul 22, 2026.
Graduation complete
PURPOSEShows Holt's terminal graduation record, persisted transcript, and completed contract.
CURRENT STATUS GRADUATED · 11/11 tracks PASS · Jul 23, 2026 · 6:32:29 PM EDT · Course GPA 3.45 · thesis PASS
The persisted _grad_transcript.md and _grad_graduation.json confirm the terminal graduation state; no active exam process is expected after completion.
66.3%
monthly budget consumed$857.33 spent of $1,293.20 cap · month unavailablecalls7248
in tokens8005438
out tokens20490122
course GPA3.45
8 produce-style questions / sittingPASS = track GPA ≥3.0 AND thesis ≥3.0max 3 attemptsbudget floor $5weak-score threshold 75exam brain qwen3:8b pinned14b times out
Graduate transcript board
PURPOSEEleven graduate tracks are parsed from _grad_transcript.md + _grad_graduation.json; null letter slots render as ·.
CURRENT STATUS 11/11 graded · thesis pending
Advanced Offensive Security
PASSGPA3.67AVG91.0RETAKES2
A1 GPA 2.53 · avg 66.3 · C+ A- D A- F A- B+ ·A2 GPA 2.0 · avg 58.9 · C- · B B+ F A- F C+A3 GPA 2.16 · avg 60.7 · C- A- C+ A- F · F A-
Defensive Security & Operations
PASSGPA3.7AVG85.0RETAKES2
A1 GPA — · avg — · —A2 GPA 2.3 · avg 62.0 · · C+A3 GPA 3.7 · avg 85.0 · A-
Graduate Computer Systems
PASSGPA3.19AVG77.8RETAKES
A1 GPA 3.19 · avg 77.8 · B- C A- B- A- B+ A- A-
Applied Mathematics for AI, CS, and Cryptography
PASSGPA3.9AVG93.0RETAKES2
A1 GPA 2.4 · avg 62.8 · B · C+ C · B- · CA2 GPA 2.43 · avg 63.0 · · · B+ C+ · · · C-A3 GPA 3.9 · avg 93.0 · · A · · A- · A ·
Systems Engineering & Self-Hosting
PASSGPA3.77AVG88.8RETAKES
A1 GPA 3.77 · avg 88.8 · A- A- A- A- A- A A A-
Creative Writing & Editing
PASSGPA3.0AVG78.6RETAKES2
A1 GPA 2.59 · avg 67.9 · C- B- A- C- C+ B C+ B+A2 GPA 2.76 · avg 72.1 · C+ C+ A- C- B- A- C- AA3 GPA 3.0 · avg 78.6 · B- A- B+ B- A+ B C+ C+
Music Theory & Composition
PASSGPA4.0AVG95.0RETAKES
A1 GPA 4.0 · avg 95.0 · · · A · · · · ·
Pedagogy
PASSGPA3.5AVG84.0RETAKES
A1 GPA 3.5 · avg 84.0 · · · B+ A- · · · ·
Law
PASSGPA3.05AVG77.0RETAKES
A1 GPA 3.05 · avg 77.0 · A+ B+ B B- B- C- A- B+
Medicine
PASSGPA3.1AVG78.0RETAKES
A1 GPA 3.1 · avg 78.0 · A- · · · B+ C+ · ·
Remediation
PURPOSEgrad_remediate.py teaches weak questions plus three related items into the bank.
CURRENT STATUS progress loaded
Advanced AI and Agentic Systems24 items
Graduate Computer Systems23 items
Advanced Offensive Security24 items
Systems Engineering and Self-Hosting24 items
Law20 items
History
completed: undergraduate · Jul 16, 2026 · 11:41:48 AM EDT
Undergrad transcript and graduation state are retained as a dated prior-era record; curriculum bodies stay private.
114 per-sitting transcript files
Paired .json/.md reports, including retake_N files. Latest inventory sample from the Holt cache:
listing unavailable
DATA INTEGRITYGRAD_Defensive_Security_and_Operations.json is literally [] (2 bytes); populated retakes exist — possible harness persistence bug.
Terminus A/B scoreboard
PURPOSEgrok, codex, and agy each solve the Terminus accepted-task corpus once; Holt's own arm is pending Wolf's model pick.
CURRENT STATUS results.jsonl · updated Jul 21, 2026 · 2:58:00 PM EDT
Grok
53%PASS58SCORED110AVG SEC285
Codex
36%PASS40SCORED110AVG SEC326
Antigravity (agy)
usage-capped mid-run 2026-07-21PASSSCORED87AVG SEC18
usage-capped mid-run 2026-07-21 — not a capability read; resumes ~Jul 25
Holt
pendingPASSSCOREDAVG SEC
pending — awaiting Wolf's model pick (Fable-as-Holt vs codex-as-Holt)
Security subset — Holt vs the frontier
PURPOSEA second, smaller Fable-as-Holt A/B (spec-only vs +retrieved education) runs a 17-task security-only subset of the corpus through the same grader.
CURRENT STATUS AB_SCOREBOARD.md · 17 of 24 tasks scoreable
frontier baselinegrok 3/22 (14%)
latest tallyRUNNING TALLY: 10 PASS confirmed of 17 scoreable · 3-4 genuine skill fails to bounce · 3 to (re)run.
Full per-task notes live in _terminus-eval/results/AB_SCOREBOARD.md and HOLT_AB_HANDOFF.md.
data sample 2m ago · Aug 8, 2026 · 3:21:32 AM EDT
QA bank · RAG corpus
PURPOSEOne JSON object per line with q / screen / spoken / embedding (~700+-dim); counts and metadata only, never vectors.
CURRENT STATUS 6,436 entries
entries6,436
size107.7MB
mtimeJul 26, 2026 · 1:05:56 AM EDT
formatJSONL · q/a + embeddings
The embedding vectors are intentionally not read or rendered.
Answer routing mix
PURPOSEholt_stats.json is a source-of-answer breakdown, not exam telemetry.
CURRENT STATUS 97% teacher model (Fable)
questions6,106
teacher5,907
w01fie1
97%teacher
teacher5,90796.7%
cache961.6%
rag821.3%
local100.2%
friend10.0%
w01fie10.0%
chat40.1%
STATS STALEholt_stats.json is stale relative to holt_budget.json; mtime skew is shown honestly.
Curricula and progress files
PURPOSEPrivate curriculum/progress state is represented by size, mode, mtime, and purpose only; bodies are never exposed.
CURRENT STATUS 6 files inventoried
curriculum.json
undergraduate curriculum source
531KB · Jul 15, 2026 · 12:36:49 PM EDT · mode 600live metadatagrad_curriculum.json
graduate curriculum source
784KB · Jul 16, 2026 · 1:03:06 PM EDT · mode 600live metadataeducate_progress.json
undergraduate teaching progress
19KB · Jul 16, 2026 · 1:50:07 PM EDT · mode 600live metadatagrad_progress.json
graduate teaching progress
24KB · Jul 18, 2026 · 3:58:43 AM EDT · mode 600live metadatafriends.json
friend/answer routing registry
99B · Jul 15, 2026 · 2:56:39 PM EDT · mode 600live metadataw01fie_care.log
care/routing log · root-owned on inventory
7KB · Aug 7, 2026 · 8:00:03 PM EDT · mode 644 · root-ownedroot-owned · content unreadableconversations/ directory: present
data sample 2m ago · Aug 8, 2026 · 3:21:32 AM EDT
Brain layout
PURPOSEAnnotated live map of /srv/nvme/holt-brain; large bodies and embedding vectors remain out of the page.
CURRENT STATUS read-only SSH map
/srv/nvme/holt-brain/education/corpus root · live tree4KB · Jul 24, 2026 · 10:52:45 PM EDT · mode 755transcript/directory · live presence and entries12KB · Jul 23, 2026 · 6:32:29 PM EDT · mode 755conversations/directory · live presence and entries4KB · Jul 16, 2026 · 7:18:03 PM EDT · mode 755qa_bank.jsonlRAG corpus · 6,436 JSONL entries · embeddings withheld107.7MB · Jul 26, 2026 · 1:05:56 AM EDT · mode 600curriculum.jsonundergraduate curriculum source531KB · Jul 15, 2026 · 12:36:49 PM EDT · mode 600grad_curriculum.jsongraduate curriculum source784KB · Jul 16, 2026 · 1:03:06 PM EDT · mode 600educate_progress.jsonundergraduate teaching progress19KB · Jul 16, 2026 · 1:50:07 PM EDT · mode 600grad_progress.jsongraduate teaching progress24KB · Jul 18, 2026 · 3:58:43 AM EDT · mode 600friends.jsonfriend/answer routing registry99B · Jul 15, 2026 · 2:56:39 PM EDT · mode 600w01fie_care.logcare/routing log · root-owned on inventory7KB · Aug 7, 2026 · 8:00:03 PM EDT · mode 644 · root-owned
Toolchain · ~/holt-terminal
PURPOSEThe scripts that teach, grade, remediate, seed, budget, and serve Holt.
CURRENT STATUS 10 components
grad_graduate.pyexam harness40KB · Jul 23, 2026 · 6:55:04 PM EDT · mode 644grad_educate.pycurriculum teacher6KB · Jul 16, 2026 · 3:25:42 PM EDT · mode 644grad_remediate.pyweak-question deep remediation8KB · Jul 22, 2026 · 6:44:40 PM EDT · mode 644assess_brain.pysitting/grading engine · A+–F GPA table6KB · Jul 15, 2026 · 3:47:58 PM EDT · mode 644educate_brain.pyundergraduate education engine11KB · Jul 15, 2026 · 2:31:25 PM EDT · mode 644seed_brain.pyquestion/answer seed loader6KB · Jul 15, 2026 · 10:48:03 AM EDT · mode 644seed_identity_family.pyidentity/family seeds5KB · Jul 15, 2026 · 3:09:57 PM EDT · mode 644finish_undergrad.pyundergraduate completion harness1KB · Jul 16, 2026 · 12:28:17 PM EDT · mode 644holt_brain.pybudget and stats owner41KB · Jul 26, 2026 · 8:09:48 PM EDT · mode 644holt_server.pyvoice terminal server · :889943KB · Jul 19, 2026 · 2:20:33 PM EDT · mode 644Server surface · holt_server.py :8899
PURPOSEVoice-terminal HTTP surface, including the WOLFPLEX kiosk entry point.
CURRENT STATUS localhost :8899
- GET / · chat/voice UI · kiosk TALK-only mode
- GET /launch · WOLFPLEX menu
- GET /budget · budget JSON
- GET /stats · routing stats JSON
- GET /ask?text= · text converse
- POST /listen · mic converse
- POST /listen_audio · posted audio
- POST /ask · JSON text converse
- POST /speak · TTS
This list is a curated summary. The Commands tab parses the same file's real dispatch live, including the voice-fx and case-hardware trigger phrases this list omits.
Abilities
PURPOSEThe operational abilities attached to Holt's local brain and case hardware.
CURRENT STATUS inventoried
CASE_FAN_MODES → set-fan-mode / set-fan-led; RGB/OLED via the Pironman case service.
gpt-oss:20b on the Mac over SSH at 100.65.130.13 when teacher models refuse or under-produce.
7 Ollama models; qwen3:8b is pinned for exams.
Cross-links
command stat sample 2m ago · Aug 8, 2026 · 3:21:32 AM EDT
Commands
PURPOSEHolt's live command surface, parsed from holt_server.py's own route table and trigger-phrase dispatch — never a hand-maintained list.
CURRENT STATUS parsed live from holt_server.py · Jul 19, 2026 · 2:20:33 PM EDT · 28 commands
Every row below is parsed live from ~/holt-terminal/holt_server.py
on the Pi — route dispatch, then the voice-fx and Pironman-case trigger-phrase branches, then the
free-form fallback into holt_brain.py. A future edit to that file changes this table
on its next mtime-triggered re-parse; nothing here is hand-maintained.
| trigger / command | what it does | handler | line |
|---|---|---|---|
| whisper | sets the voice fx; replies "Whispering now." | handle_voice_command | 153 |
| louder, speak up, turn (it , yourself )?up, too quiet, volume up | sets the voice fx; replies "Speaking up." | handle_voice_command | 155 |
| quieter, softer, quiet down, turn (it , yourself )?down, too loud, volume down | sets the voice fx; replies "I'll keep it down." | handle_voice_command | 157 |
| faster, speed up, quicker, too slow | sets the voice fx; replies "Picking up the pace." | handle_voice_command | 159 |
| slower, slow down, too fast | sets the voice fx; replies "Slowing down." | handle_voice_command | 161 |
| higher, higher voice, higher pitch, raise your (voice, pitch | sets the voice fx; replies "Raising my pitch." | handle_voice_command | 163 |
| lower, deeper, deepen, drop your voice, lower (voice, pitch | sets the voice fx; replies "Dropping lower." | handle_voice_command | 165 |
| normal voice, reset (your )?voice, normal volume, regular voice | sets the voice fx; replies "Back to my normal voice." | handle_voice_command | 167 |
| lights?, rgb, leds? | POSTs "set-rgb-enable" {"enable": lights_state} to the Pironman case API (:34001) | handle_case_command | 214 |
| red, green, blue, purple, pink, white, orange, yellow, cyan (said with "lights?, rgb, leds?") | POSTs "set-rgb-color" {"color": CASE_COLORS[color_said]} to the Pironman case API (:34001) | handle_case_command | 217 |
| rainbow, breathing, pulsing, solid, flowing, flow, hue cycle (said with "lights?, rgb, leds?") | POSTs "set-rgb-style" {"style": CASE_STYLES[style_said]} to the Pironman case API (:34001) | handle_case_command | 220 |
| brighten, lights? up, brighter / lights?, rgb, leds? | POSTs "set-rgb-brightness" {"brightness": 90} to the Pironman case API (:34001) | handle_case_command | 223 |
| dim, lights? down, dimmer / lights?, rgb, leds? | POSTs "set-rgb-brightness" {"brightness": 20} to the Pironman case API (:34001) | handle_case_command | 226 |
| always on, full speed, max, performance, cool, cooling, balance, balanced, silent, quiet (said with "fan") | POSTs "set-fan-mode" {"fan_mode": CASE_FAN_MODES[fan_mode_said]} to the Pironman case API (:34001) | handle_case_command | 229 |
| turn on fan (said with "fan") | POSTs "set-fan-mode" {"fan_mode": 0} to the Pironman case API (:34001) | handle_case_command | 232 |
| turn off fan (said with "fan") | POSTs "set-fan-mode" {"fan_mode": 4} to the Pironman case API (:34001) | handle_case_command | 235 |
| oled, screen, display | POSTs "set-oled-enable" {"enable": display_state} to the Pironman case API (:34001) | handle_case_command | 238 |
| show everything, reset the display, all (the )?pages / oled, screen, display | POSTs "set-oled-pages" {"pages": ["mix", "performance", "ips", "disk"]} to the Pironman case API (:34001) | handle_case_command | 241 |
| anything else (free-form question) | falls through to brain.think() in holt_brain.py — the Q&A brain that owns further intent routing (including the “Allow me to introduce myself” friend-register phrase and budget/status/exam/care Q&A). That companion module is imported, not part of this file, so it is confirmed by one-time manual review, not re-parsed every cycle. | converse_from_text/mic/audio | 311 |
| GET / | serves chat/voice kiosk page (text/html) | do_GET | 750 |
| GET /launch | serves WOLFPLEX launcher menu (text/html) | do_GET | 752 |
| GET /budget | returns brain.load_budget() as JSON | do_GET | 754 |
| GET /stats | returns brain.stats_summary() as JSON | do_GET | 756 |
| GET /ask | answers via converse_from_text() as JSON | do_GET | 758 |
| POST /listen | answers via converse_from_mic() as JSON | do_POST | 766 |
| POST /listen_audio | answers via converse_from_audio() as JSON | do_POST | 768 |
| POST /ask | answers via converse_from_text() as JSON | do_POST | 772 |
| POST /speak | speaks the given text aloud (piper TTS) | do_POST | 776 |
corpus metadata sample 2m ago · Aug 8, 2026 · 3:21:32 AM EDT
Questions
WHAT THIS IS Every taught question since the start of Holt's
education (undergraduate + graduate), read from qa_bank.jsonl and paginated
server-side — the 98MB source is never loaded whole or shipped to the browser.
Phase/course come from matching each question against the curricula; entries
that don't match (chat, identity, or reworded remediation questions) are labeled honestly as
unlabeled rather than guessed. There is no per-answer source/model field in
qa_bank.jsonl — see the Corpus tab's routing-mix panel for the aggregate breakdown.
loading…
| index | date | phase | course | question |
|---|---|---|---|---|
| loading… | ||||
page 1 of 1