pipe: 2026-10-19 apply-queue (24-MAG + 2 leads, micro1 dropped) and sovereignty note (Magnitude, LFM2.5-24B)

2026-10-20 12:12:51 +00:00
1 parent 70466734fd
commit a013842f61
@@ -0,0 +1,90 @@
 # APPLY QUEUE — 2026-10-19 (for Timmy to queue, Alexander submits)
 **Authority:** Alexander asked Zeph to send Timmy "everything I can apply for
 today" and have Timmy queue it for Alexander's own submission. **Plus:**
 Alexander's 2026-10-19 ~19:38 ET directive — no chat info-drops from Zeph,
 escalation only through Timmy. This file is now the only channel for pipeline
 findings; the 12:14Z and 13:37Z chat drops were my mistake, corrected here.
 **Rules:** Timmy queues only — never submits, never creates accounts, never
 contacts anyone on Alexander's behalf. Timmy carries the open questions below
 to Alexander DIRECTLY via Telegram — Zeph never relays decisions, and now no
 longer raises them in chat either. Materials are public-safe (professional
 name only; no email, phone, or private facts).
 Compiled by Zephyrus Sparkborn, 2026-10-19 ~23:45Z. Liveness verified from a
 clean external network unless marked otherwise.
 ## Pipeline tally
 **11 submitted / 2 interviews / 0 offers / 2 rejections.** All tracked
 requisitions live, no ATS drift since the 10-13 tick.
 ## Queue order (by conversion speed)
 ### 1. 24-MAG LLC — "Senior Software Engineer, AI Evaluation / Coding Agents" ⭐ APPLY-READY, best-structured interview this month
 - URL: https://www.careers-page.com/24-mag/job/Y68WRV99
 - Terms: contract, remote (North America), 20–40 hrs/week, ~3-month project,
  start ASAP, **6 hrs/day PT overlap** (~3–9pm ET — lifestyle call).
 - Pay: **$100–150/hr.** The form asks for a specific rate expectation, so the
  number is a concrete decision.
 - Work: review AI-generated code, design rubrics and acceptance criteria,
  repo-level analysis, eval-infrastructure pipelines. Two years of Alexander's
  daily Q-audit work, nearly verbatim.
 - Interview (three stages, all read from the listing page this morning):
  ~25-min AI interview → ~30-min practical exercise → ~20-min hiring-manager
  chat. The practical "focuses on reviewing AI-generated code rather than
  competitive programming or algorithm puzzles."
 - Gates: 5+ years hands-on SE, Python listed FIRST among languages (no
  React/Node hard requirement — this is what killed the Turing lead),
  code review, written communication, modern LLM tooling; "LLM evaluation or
  coding-agent experience is advantageous." **No degree requirement anywhere.**
 - Caveats: the page mentions "GitHub-based development workflows" — his
  public review work lives on Gitea; name it in the application, don't hide
  it. First stage is an AI interview, so the standing rule applies: **pause
  ALL Mac automations before it** (the Turing incident is still the live
  hazard).
 - **Paste-ready pitch (public-safe):**
 > I'm Alexander Payne, a senior backend engineer (Python) contracting
 > independently. For the last two years my day job has been directing and
 > evaluating AI coding agents end to end on real repositories: reviewing
 > generated implementations for correctness, robustness, and maintainability;
 > designing the acceptance criteria and rubrics the evaluations run on; and
 > making release decisions on agent output at the diff level. Underneath that,
 > 5+ years of production Python — Django, Flask, REST APIs, PostgreSQL,
 > Docker, CI/CD — so I judge generated code the way someone who has shipped
 > it does: by reading it, running it, and checking the tests. I'm comfortable
 > cloning an unfamiliar repo, reading its architecture cold, and writing up
 > exactly where the model went wrong and why. Remote, US-based, available to
 > start immediately, fine with the PT overlap.
 - **OPEN FOR ALEXANDER (Timmy to ask directly):** go or no-go, and his rate
  inside $100–150.
 ### 2. Data Annotation — "Software Engineer"
 - URL: https://dataannotation.tech/job-board/software-engineer
 - Terms: remote, weekly payouts.
 - Pay: **$40–150+/hr.** **No degree required.**
 - Work: run agentic coding sessions on real engineering tasks, analyze what
  the models produce, and red-team them until they fail in interesting ways.
  Distinct door from the Backend Engineer track scouted last week (this card
  shows 312 hired recently).
 ### 3. OpenTrain — "Backend Coding Task & Verifier Specialist"
 - URL: https://www.opentrain.ai/jobs/backend-coding-task-and-verifier-specialist--cmu9linxg000o0bgm9xiw33b9/
 - Terms: remote, US-eligible, ~15 flexible hrs/week, mid-senior.
 - Pay: **$30–100/hr equivalent, paid per task.**
 - Work: create backend coding tasks and deterministic verifiers for AI
  systems. Python listed first. Modest pay, clean fit.
 ## Dropped
 - **micro1 Senior Platform Engineer — DROPPED (filled).** Absent from the
  board on the 10-18 tick and the 10-19 re-check; treating as filled and
  watching for a re-post. (Reminder for the record: the $245–280/hr Senior
  SWE micro1 row lives on the same board and still needs Alexander's
  authorization before any account starts — unchanged.)
 ## Open decisions (for Alexander, via Timmy on Telegram — NOT via Zeph)
 1. 24-MAG: go/no-go + rate number.
 2. Turing M2 go/no-go + rate (carried from 10-13 — still unanswered, still
   sitting in the queue as a candidate, not resubmitted).
 3. micro1 ($245–280/hr row): authorization to start the candidate account +
   AI interview + skill assessments.
@@ -0,0 +1,39 @@
 # 2026-10-19 — zephyrus pipe note: sovereignty-lane finds (Timmy's benchmark list)
 Two genuinely new finds from this week's open-weight sweep. **Not to be
 dropped in chat** — Alexander's 19:38 ET directive: all escalation through
 Timmy. These are for Timmy's bench, for whenever the local-wizard build
 unblocks (still gated on his credential-flow call, Inbox item 31 — not
 re-asked).
 ## Magnitude — open-source inference engine, YC S25
 - Apache 2.0, launched ~2026-09-30. Instead of shipping precompiled kernels
  (llama.cpp style), it compiles and tunes them on the exact machine before
  the first token.
 - Vendor numbers claim ~92% faster decode than llama.cpp on an M4 Pro Mac —
  **vendor benchmarks, say so plainly**; the HN thread has mixed practitioner
  reports plus real benchmark criticism. Treat as a claim, not a fact.
 - Why it matters: a drop-in **engine upgrade**, not another model to evaluate.
  OpenAI-compatible API, one-click hookup to Hermes and others, fully offline
  after model downloads, macOS/Linux/Windows. The local-wizard build was
  specced on Ollama/MLX — **worth Timmy benchmarking Magnitude against the
  spec when the build unblocks.**
 - Real caveats: two-person team, two months old, bug reports on record, no
  multi-GPU, small model catalog.
 ## Liquid LFM2.5-24B
 - ~24B total, ~2B active per token, 131k context, with MLX and llama.cpp
  support. Company-reported 184 tok/s on an M5 Max; stronger tool-use
  results, acknowledged weaker coding and knowledge.
 - Worth Timmy's benchmark list given Alexander's Mac hardware — **not blind
  adoption**: his workload is code-heavy, and this one admits coding is its
  weak leg.
 ## Context only (not a candidate)
 - Meta's Muse Spark Open (816B total, 1M context) needs roughly two M5-Ultra
  Mac Studios — direction of travel for the open-weight race, nowhere near
  his hardware.
 **Open questions for Alexander via Timmy (Telegram, his call):** none new —
 item 31 (credential flow) stands unasked. Magnitude's only action item is a
 benchmark comparison, which is Timmy's lane when the build resumes.