Archived dispatch

What are the key findings in "The Frontier AEO Tracker: What Astra Chooses (and every other frontier model, and what you can do about it)"?

Lowconfidence4 sub-claims remain below the evidence threshold

9/13/2026, 12:33:51 AM · llm:deepseek:deepseek-v4-flash

The dispatch, itemised.

§ IThe decision$0.019 / $0.03
63%$0.011 under cap
Decompose

Breaking down: "What are the key findings in "The Frontier AEO Tracker: What Astra Chooses (and every other frontier model, and what you can do about it)"?"

Decompose

Identified 4 research target(s) to investigate; these are not established facts

Decompose

Deep mode: up to 4 paid/cached reads plus one bounded gap-expansion pass when needed.

Discover

Discovered 21 verified source(s)

Discover

Recalled 40 past runs on this subject — how these sources performed when they were available.

Discover

ERC-8004 reputation loaded — composite scores on this subject.

Pre-check

Claim-aware portfolio selected 1/2 positive proposal(s): 0 cached + 1 fresh, predicting 4/4 claim(s) above the evidence floor with $0.004000/$0.015000 fetch USDC reserved.

Pre-check

Free-preview pre-check maps an actionable source to every sub-claim (4/4); paid reading may proceed within the budget.

DecideBUY
Latent.Space — The Frontier AEO Tracker: What Astra Chooses (and every other frontier model, and what you can do about it)$0.004 · EV 95%

This is the exact article named in the question — Latent.Space's 'The Frontier AEO Tracker: What Astra Chooses (and every other frontier model, and what you can do about it)'. It is the only candidate whose title matches the target, it is full_text (4256 bytes) and uncached, so it can directly answer all four sub-claims: the tracker's key findings, what Astra chooses, what other frontier models choose, and the recommended actions. Latent.Space has a 50% citation rate on this subject, and no other source covers this specific AEO tracker. Price $0.004 is well within budget. — selected for the claim-aware evidence portfolio (targets claims 1, 2, 3, 4; $0.004000 fetch USDC, 1 attention slot).

DecideSKIP
Decrypt — GPT-6 Astra Users Say OpenAI's Newest Model Got Dumber. It Happened Before, Too$0.002 · EV 35%

Decrypt's piece on GPT-6 Astra users complaining the model was 'nerfed' is the only other candidate that mentions Astra by name, so it could corroborate or contextualize what the tracker says about Astra's behavior (sub-claim 1). However it is a narrow abstract (148 bytes) about user complaints, not the AEO tracker itself, and Decrypt has only a 25% citation rate here. Cheap at $0.002, worth a look as secondary color. — the claim-aware portfolio chose a stronger, less redundant set inside the 4-source attention and $0.015000 fetch-budget caps, so this proposal stays unspent.

DecideSKIP
Simon Willison's Weblog — Anthropic’s best AI model struggles to attract users as cheaper tools thrive$0.003 · EV 15%

Simon Willison's post on Anthropic's model struggling against cheaper tools is topically adjacent (frontier model competition) but deliveryKind is metadata_only with 0 plaintext bytes — no actual content to read, and the title gives no findings about the AEO tracker or Astra's choices. Not worth $0.003.

DecideSKIP
Hugging Face - Blog — Give Your Coding Agents a Memory You Own$0.003 · EV 10%

Hugging Face's 'Give Your Coding Agents a Memory You Own' is about agent memory tooling, unrelated to the Frontier AEO Tracker's findings on model choices. It is also metadata_only with 0 plaintext bytes, so nothing can be read. Skip.

DecideSKIP
Conzit Labs — Building a Transparent Language Model in Node.js$0.002 · EV 10%

Conzit Labs has a strong 75% citation rate on this subject, but this particular article is about building a transparent language model in Node.js — an implementation tutorial, not the AEO tracker or frontier-model selection behavior. The 153-byte abstract shows no overlap with any sub-claim.

DecideSKIP
Cointelegraph.com News — Crypto firms still seeking frontier AI access; only select few have it$0.002 · EV 20%

Cointelegraph's piece on crypto firms seeking frontier AI access touches frontier-model access policy, which is loosely related to 'what every other frontier model chooses,' but the 157-byte abstract is about crypto firms' access restrictions, not the AEO tracker's findings. Too tangential to justify a read.

DecideSKIP
Stripe Blog — What Link data tells us about AI spending$0.002 · EV 10%

Stripe's Link data on AI spending is about consumer payment patterns, not AEO tracking or frontier model choices. No sub-claim is supported by the abstract.

DecideSKIP
Ethereum Foundation Blog — The triage is the product: running AI agents against Ethereum's protocol code$0.002 · EV 10%

Ethereum Foundation's post on running AI agents against protocol code is about security triage workflows, unrelated to the Frontier AEO Tracker's findings about Astra and other frontier models.

DecideSKIP
Stablecoin Ledger — Why USDC settles instantly onchain$0.003 · EV 5%

Stablecoin Ledger's USDC settlement piece is off-topic for an AEO/frontier-model tracker question, and this source has never been cited on this subject (0/1). No sub-claim overlap.

DecideSKIP
Agent Economy Weekly — x402 turns HTTP 402 into an agent payment rail$0.004 · EV 5%

Agent Economy Weekly's x402 payment-rail article concerns agent payments, not the AEO tracker or model selection behavior. No relevant targets.

DecideSKIP
Onchain Micropayments Digest — Nanopayments and the $0.000001 floor$0.005 · EV 5%

Onchain Micropayments Digest covers nanopayment floors and batching — unrelated to the AEO tracker question.

DecideSKIP
Distributed Systems Notes — Idempotency keys prevent double-spends$0.003 · EV 5%

Distributed Systems Notes on idempotency keys is a database/consensus topic with no bearing on the AEO tracker's findings.

DecideSKIP
Garden & Soil Monthly — Building a no-dig raised bed$0.002 · EV 0%

Gardening article — entirely off-topic for a frontier-model AEO tracker question.

DecideSKIP
Retro Game Hardware — Recapping a 1990s console$0.002 · EV 0%

Retro console recapping — entirely off-topic.

DecideSKIP
The Coinbase Blog - Medium — In response to the Wall Street Journal$0.003 · EV 5%

Coinbase's response to the WSJ concerns proprietary trading allegations, not AEO tracking or frontier model choices. Despite Coinbase's strong reputation, the content is irrelevant here.

DecideSKIP
CoinDesk: Bitcoin, Ethereum, Crypto News and Price Data — Crypto Long & Short: Inside the 300-to-1 onchain gap between the dollar and euro$0.002 · EV 5%

CoinDesk's dollar-vs-euro onchain gap piece is stablecoin market analysis, unrelated to the AEO tracker question.

DecideSKIP
Inner Axiom — The Codex — The Pleiades, the Seven Sisters in Taurus and Orion$0.002 · EV 0%

Inner Axiom's Pleiades mythology article is off-topic, and this source has the weakest reputation on this subject (18%).

DecideSKIP
Vitalik Buterin's website — Low-risk defi can be for Ethereum what search was for Google$0.004 · EV 5%

Vitalik's low-risk DeFi essay is about Ethereum economics, not AEO tracking; it is also metadata_only with 0 plaintext bytes.

DecideSKIP
Arc Settlement Benchmarks — Measuring x402 settlement latency on Arc$0.003 · EV 5%

Arc settlement latency benchmarks concern x402 payment finality, unrelated to the AEO tracker's findings on model choices.

DecideSKIP
Web Payments Review — How long do x402 payments take to finalize?$0.002 · EV 5%

Web Payments Review's x402 finalization timing is a payments topic with no overlap with the AEO tracker question.

DecideSKIP
Keryx Engineering (first-party) — Recovering a Keryx paid research job$0.002 · EV 10%

Keryx's first-party buyer-recovery notes are about payment recovery mechanics, not the Frontier AEO Tracker's findings. High reputation but wrong topic for this question.

Fetch

Agent spend wallet ready: 0x29028Fe1122E17Fe7863A22701e863FE4DaE1aFB (balance sufficient)

Fetch

Paying $0.004 toll to read Latent.Space — The Frontier AEO Tracker: What Astra Chooses (and every other frontier model, and what you can do about it)…

Fetch

Paid $0.004 to Latent.Space — The Frontier AEO Tracker: What Astra Chooses (and every other frontier model, and what you can do about it) (settled 44cb669e-c…) — S1

Sufficiency

Sub-claim "What are the key findings reported in "The Frontier AEO Trac…": 20% covered by S1

Sufficiency

Sub-claim "What does the source say Astra chooses, according to the tra…": 10% covered by S1

Sufficiency

Sub-claim "What does the source say every other frontier model chooses,…": 10% covered by S1

Sufficiency

Sub-claim "What does the source recommend readers can do about it?": 0% covered

Sufficiency

The supplied passages from the Latent.Space article are excerpted and do not contain the tracker's actual key findings, specific choices attributed to Astra, choices attributed to every other frontier model, or recommendations for readers. Passage S1 provides only background motivation, a note about contamination/bias, mention of close contests and special reports, and a fun side feature. It does not state the requested findings or recommendations. The assessment does not establish a complete supported answer for every requested part.

Re-evaluate

Sub-claim "What are the key findings reported in "The Frontier AEO Trac…": 40% covered by S1 — S1 gives partial key findings: dominant products in categories, contamination checked, bias exists (models favor their own products), close contests as AEO battlegrounds, and consequential flips between model generations. But the excerpt is truncated and does not enumerate the actual rankings or the specific findings in full.

Re-evaluate

Sub-claim "What does the source say Astra chooses, according to the tra…": 40% covered by S1 — S1 states 'Sol/Astra like Codex,' indicating Astra chooses Codex, but no broader or detailed Astra choices are provided.

Re-evaluate

Sub-claim "What does the source say every other frontier model chooses,…": 40% covered by S1 — S1 gives examples: Fable/Opus like Claude Code, Grok loves Cursor, Muse loves Muse Code, SWE-1.7 loves Devin. This is partial and not an exhaustive account of every other frontier model's choices.

Re-evaluate

Sub-claim "What does the source recommend readers can do about it?": 10% covered by S1 — S1 mentions AEO battlegrounds, special reports, a family feud game, and openness to business enquiries, but does not state explicit recommendations for readers about what to do.

Re-evaluate

Coverage is partial across all sub-claims, but the skipped sources are unrelated to the Frontier AEO Tracker's findings (they cover GPT-6 Astra complaints, Anthropic user struggles, coding agent memory, Node.js LM transparency, crypto AI access, payments, gardening, retro hardware, etc.). None would fill the specific gaps about Astra's choices, other frontier models' choices, or reader recommendations. Buying them would not improve coverage, so no purchase is recommended.

Sufficiency

Final check — "What are the key findings reported in "The Frontier AEO Trac…": 20% assessed by S1

Sufficiency

Final check — "What does the source say Astra chooses, according to the tra…": 30% assessed by S1

Sufficiency

Final check — "What does the source say every other frontier model chooses,…": 30% assessed by S1

Sufficiency

Final check — "What does the source recommend readers can do about it?": 0% assessed

Sufficiency

Final coverage assessment — The supplied passages from the Latent.Space article provide only high-level framing and teaser text. They mention that the tracker exists, that results show dominant products, that bias exists (e.g., Fable/Opus like Claude Code, Sol/Astra like Codex, Grok loves Cursor, Muse loves Muse Code, SWE-1.7 loves Devin), and that there are close contests and special reports. However, the passages do not state the actual key findings, do not specify what Astra chooses beyond a single example, do not state what every other frontier model chooses, and do not give the recommended actions readers can take. The article is excerpted and the core findings/recommendations are missing. The assessment does not establish a complete supported answer for every requested part.

Synthesize

Synthesizing a grounded answer from 1 source(s)…

Evidence

Relevance review returned; only checked excerpts can retain support, and review cannot raise it.

Evidence

Below reward gate — S1 supports claim 1 at 0%: “Naive autoresearch investment in our AEO have yielded impressive ROI, and so naturally it was time to take it seriously.”

Evidence

Below reward gate — S1 supports claim 1 at 10%: “Basic Results Here are the most dominant products (in their categories) in the world: There are some familiar names in there — opening…”

Evidence

Verified — S1 supports claim 2 at 90%: “However, bias does exist - when models are asked for coding agent recommendations, Fable/Opus like Claude Code and Sol/Astra like Codex and …”

Evidence

Verified — S1 supports claim 3 at 90%: “However, bias does exist - when models are asked for coding agent recommendations, Fable/Opus like Claude Code and Sol/Astra like Codex and …”

Evidence

Below reward gate — S1 supports claim 4 at 20%: “See more We also made a little family feud type game where you can see if your priors align with the data.”

Synthesize

Drafted answer citing 1 source(s)

Verdict

Confidence: Low — 4 sub-claims remain below the evidence threshold.

Attribute

Latent.Space contributed 100% → reward $0.015

Settle

Settled $0.015 citation reward → Latent.Space (a3500812-f…)

Done

Done. Spent $0.019 across 2 confirmed/simulated payment(s) to creators.

§ IIIThe settlementweighted · USDC on Arc
100%
1

Latent.Space

batched

100%$0.015
§ IIThe reading1 cited
Lowconfidence4 sub-claims remain below the evidence thresholddeep researchpreview plan 4/4 claimsportfolio 1/2 · evidence 100%

> ⚠ Low confidence — 4 sub-claims remain below the evidence threshold within budget. Treat this as provisional.

The supplied passages from the article do not state a single consolidated list of "key findings." What they do report is: the tracker was built after naive autoresearch investment in AEO yielded impressive ROI , and it was inspired by "What Claude Code Actually Chooses" and then extended/adjusted to the authors' tastes . The "Basic Results" section presents the most dominant products in their categories, with familiar names raising the contamination question, which the authors say they checked . They also state that bias exists: when models are asked for coding agent recommendations, Fable/Opus like Claude Code, Sol/Astra like Codex, Grok loves Cursor, Muse loves Muse Code, and SWE-1.7 loves Devin . They note many "close contests" and "always the vibesmaid, never the vibe" categories as key AEO battlegrounds , and say they prepared special reports observing very consequential flips in model choices between model generations from the same lab .

On what Astra chooses, the only relevant statement in the supplied passages is that Sol/Astra like Codex . The passages do not otherwise specify what Astra chooses.

On what every other frontier model chooses, the passages give examples rather than an exhaustive account: Fable/Opus like Claude Code, Grok loves Cursor, Muse loves Muse Code, and SWE-1.7 loves Devin . The passages do not provide a complete list for every other frontier model.

On what readers can do about it, the supplied passages do not contain a recommendation. They mention that every prompt and answer pair is inspectable , that there is a family feud type game to see if your priors align with the data , and that the authors are open to further suggestions and business enquiries , but no explicit "what you can do" recommendation is present in the supplied excerpts.

Evidence ledger — quotes verified before rewards

  1. What are the key findings reported in "The Frontier AEO Tracker: What Astra Chooses (and every other frontier model, and what you can do about it)"?

    0%

    No reward-qualifying evidence

  2. What does the source say Astra chooses, according to the tracker?

    30%
    However, bias does exist - when models are asked for coding agent recommendations, Fable/Opus like Claude Code and Sol/Astra like Codex and Grok loves Cursor and Muse loves Muse Code and SWE-1.7 loves Devin and so on. [S1] The Frontier AEO Tracker: What Astra Chooses (and every other frontier model, and what you can do about it)
  3. What does the source say every other frontier model chooses, according to the tracker?

    30%
    However, bias does exist - when models are asked for coding agent recommendations, Fable/Opus like Claude Code and Sol/Astra like Codex and Grok loves Cursor and Muse loves Muse Code and SWE-1.7 loves Devin and so on. [S1] The Frontier AEO Tracker: What Astra Chooses (and every other frontier model, and what you can do about it)
  4. What does the source recommend readers can do about it?

    0%

    No reward-qualifying evidence

Footnotes — each one pays its author

Helpful?
Spent$0.019
To creators100%
Decisions1 bought · 0 cached · 20 skipped
llm:deepseek:deepseek-v4-flash

Portable research receipt

Take the evidence trail with you

One deterministic JSON bundle binds the answer, visible decisions, exact article versions, claim evidence and a Circle-settlement snapshot under SHA-256. Retain the digest to detect later changes; the self-check is not a publisher or Keryx signature.

Exact receipt still current

1 exact cited article version still match Keryx's current index. The source cited here has published nothing new since this dispatch settled.

Inspect machine-readable audit

Ask a follow-upNew dispatch · creators paid again

Carries this dispatch’s question as context — never its answer. The next dispatch is read from sources bought for it.

From the archive

Related dispatches