Morning Singularity Digest - 2026-07-29

Context

Feed it context -- your CV, your career story, your proof points, your preferences, what you're good at, what you want to avoid.

What's new

Heads up: the first evaluations won't be great.

Key details

• So I engineered the system I wish I had.
• Companies use AI to filter candidates.
• I just gave candidates AI to choose companies.
• FEATURED IN 740+ job listings evaluated · 100+ personalized CVs · 1 dream role landed Also runs on any agent-skill-standard CLI.

Results & evidence

• Open-source AI job search: scan job portals, evaluate listings with a structured A-F rubric into a 1.0-5.0 score, tailor your CV, track applications — runs locally in your AI coding CLI (Claude Code, Codex, OpenCode, Antigravity…) English | Español | Deutsc...
• FEATURED IN 740+ job listings evaluated · 100+ personalized CVs · 1 dream role landed Also runs on any agent-skill-standard CLI.
• Instead of manually tracking applications in a spreadsheet, you get an AI-powered pipeline that: - Evaluates offers with a structured evaluation -- blocks A-F scored across 5 weighted dimensions, plus block G, a separate posting-legitimacy assessment that n...

Limitations / unknowns

• Generalization outside curated tasks is still unclear.

Next-step validation checks

• Reproduce one claim with a public baseline and fixed evaluation settings.
• Check robustness on out-of-distribution or long-context cases.
• Track whether independent teams report matching results.

Context

Instead, you are programming the program.md Markdown files that provide context to the AI agents and set up your autonomous research org.

What's new

AI agents running research on single-GPU nanochat training automatically One day, frontier AI research used to be done by meat computers in between eating, sleeping, having other fun, and synchronizing once in a while using sound wave interconnect in the ri...

Key details

• Research is now entirely the domain of autonomous swarms of AI agents running across compute cluster megastructures in the skies.
• The agents claim that we are now in the 10,205th generation of the code base, in any case no one could tell if that's right or wrong as the "code" is now a self-modifying binary that has grown beyond human comprehension.
• This repo is the story of how it all began.
• The idea: give an AI agent a small but real LLM training setup and let it experiment autonomously overnight.

Results & evidence

• The agents claim that we are now in the 10,205th generation of the code base, in any case no one could tell if that's right or wrong as the "code" is now a self-modifying binary that has grown beyond human comprehension.
• It modifies the code, trains for 5 minutes, checks if the result improved, keeps or discards, and repeats.

Limitations / unknowns

• Generalization outside curated tasks is still unclear.

Next-step validation checks

• Reproduce one claim with a public baseline and fixed evaluation settings.
• Check robustness on out-of-distribution or long-context cases.
• Track whether independent teams report matching results.

Context

arXiv:2607.24882v1 Announce Type: cross Abstract: Modern coding agents are usually evaluated by whether they eventually produce a correct patch, but patch generation depends on an earlier context-acquisition stage: finding the repository files needed for th...

What's new

arXiv:2607.24882v1 Announce Type: cross Abstract: Modern coding agents are usually evaluated by whether they eventually produce a correct patch, but patch generation depends on an earlier context-acquisition stage: finding the repository files needed for th...

Key details

• We introduce Agent Retrieval Bench, a file-level benchmark for this upstream retrieval problem.
• Samples are built from real coding-workflow signals and evaluated against frozen base-commit repositories, with relevance defined by what an agent needs next rather than direct query-file semantic similarity.
• The benchmark covers four positive-retrieval tasks: code2test, comment2context, trace2code, and edit2ripple; a fifth subset evaluates selective retrieval using natural evidence-backed no-gold cases and counterfactual wrong-repository controls.
• Agent Retrieval Bench contains 427 samples across 25 repositories: 345 positive examples, 50 natural no-gold examples, and 32 counterfactual controls.

Results & evidence

• arXiv:2607.24882v1 Announce Type: cross Abstract: Modern coding agents are usually evaluated by whether they eventually produce a correct patch, but patch generation depends on an earlier context-acquisition stage: finding the repository files needed for th...
• Agent Retrieval Bench contains 427 samples across 25 repositories: 345 positive examples, 50 natural no-gold examples, and 32 counterfactual controls.
• The corpus includes 308 base-commit snapshots, 392,000 files, and 7.9 million chunks.

Limitations / unknowns

• Generalization outside curated tasks is still unclear.

Next-step validation checks

• Reproduce one claim with a public baseline and fixed evaluation settings.
• Check robustness on out-of-distribution or long-context cases.
• Track whether independent teams report matching results.

Context

We show the mismatch to be an evaluation problem in the costume of a training one at a fixed recall, prior correction and calibration cannot move precision, and answer it with a prior-matched reporting method based on three figures: balanced-test, operation...

What's new

We show the mismatch to be an evaluation problem in the costume of a training one at a fixed recall, prior correction and calibration cannot move precision, and answer it with a prior-matched reporting method based on three figures: balanced-test, operation...

Key details

• Because attention is the cost of error, precision leads.
• Its classifier was trained and reported at a one-to-one class balance, fixed before the operational rate could be known.
• That rate has since emerged at roughly one scene in twenty, and a balanced-test score badly overstates the precision a validator meets.
• A model that scores 0.794 balanced-test precision scores 0.192 in real operation: the gap is a systematic artefact of reporting at the wrong prior, invisible to the metric most work quotes.

Results & evidence

• arXiv:2607.07146v2 Announce Type: replace Abstract: The Internal Waves Service screens the Sentinel-1 Wave-mode archive for internal solitary waves, routing detections to experts whose adjudication time is the resource the effort exists to conserve.
• A model that scores 0.794 balanced-test precision scores 0.192 in real operation: the gap is a systematic artefact of reporting at the wrong prior, invisible to the metric most work quotes.
• Holding recall at a floor of 0.80 and certifying against a sealed, single-read lockbox, the promoted model reports 0.927 precision at the operational prior; an out-of-time check confirms discrimination transfers to unseen periods while a fixed operating poi...

Limitations / unknowns

• Generalization outside curated tasks is still unclear.

Next-step validation checks

• Reproduce one claim with a public baseline and fixed evaluation settings.
• Check robustness on out-of-distribution or long-context cases.
• Track whether independent teams report matching results.

Context

You seed baits, artifacts crafted to look exactly like the high-value, security-relevant assets such an agent is hunting for, staged in the places it will look and wrapped in the context that sells them: a plausible filename, a companion ciphertext blob, a...

What's new

by Cracken; Core team: Dario Pasquini and Michal Bazyli Blacksea is an active honeypot and canary-bait control system built to detect and drown LLM-driven attackers: autonomous AI agents and LLM-assisted operators that scan and exploit systems.

Key details

• Blacksea doesn't stop at watching LLM attacks.
• It exploits flaws in the attacker's LLM judgment to gain arbitrary code execution on their machines, collect intel passive defenses can't reach, and make sure they don't come back.
• An LLM-driven attacker works an engagement by reasoning toward the assets that move it forward: credentials to reuse, a decryptor for an encrypted blob, a key-derivation tool, a config unpacker, a token minter, an internal API client.
• Blacksea turns that reasoning into a trap.

Results & evidence

• No hard numbers surfaced in the source text; treat claims as directional until benchmarks appear.

Limitations / unknowns

• Generalization outside curated tasks is still unclear.

Next-step validation checks

• Reproduce one claim with a public baseline and fixed evaluation settings.
• Check robustness on out-of-distribution or long-context cases.
• Track whether independent teams report matching results.

Context

Feed it context -- your CV, your career story, your proof points, your preferences, what you're good at, what you want to avoid.

What's new

Heads up: the first evaluations won't be great.

Key details

• So I engineered the system I wish I had.
• Companies use AI to filter candidates.
• I just gave candidates AI to choose companies.
• FEATURED IN 740+ job listings evaluated · 100+ personalized CVs · 1 dream role landed Also runs on any agent-skill-standard CLI.

Results & evidence

• Open-source AI job search: scan job portals, evaluate listings with a structured A-F rubric into a 1.0-5.0 score, tailor your CV, track applications — runs locally in your AI coding CLI (Claude Code, Codex, OpenCode, Antigravity…) English | Español | Deutsc...
• FEATURED IN 740+ job listings evaluated · 100+ personalized CVs · 1 dream role landed Also runs on any agent-skill-standard CLI.
• Instead of manually tracking applications in a spreadsheet, you get an AI-powered pipeline that: - Evaluates offers with a structured evaluation -- blocks A-F scored across 5 weighted dimensions, plus block G, a separate posting-legitimacy assessment that n...

Limitations / unknowns

• Generalization outside curated tasks is still unclear.

Next-step validation checks

• Reproduce one claim with a public baseline and fixed evaluation settings.
• Check robustness on out-of-distribution or long-context cases.
• Track whether independent teams report matching results.

Context

arXiv:2607.24882v1 Announce Type: cross Abstract: Modern coding agents are usually evaluated by whether they eventually produce a correct patch, but patch generation depends on an earlier context-acquisition stage: finding the repository files needed for th...

What's new

arXiv:2607.24882v1 Announce Type: cross Abstract: Modern coding agents are usually evaluated by whether they eventually produce a correct patch, but patch generation depends on an earlier context-acquisition stage: finding the repository files needed for th...

Key details

• We introduce Agent Retrieval Bench, a file-level benchmark for this upstream retrieval problem.
• Samples are built from real coding-workflow signals and evaluated against frozen base-commit repositories, with relevance defined by what an agent needs next rather than direct query-file semantic similarity.
• The benchmark covers four positive-retrieval tasks: code2test, comment2context, trace2code, and edit2ripple; a fifth subset evaluates selective retrieval using natural evidence-backed no-gold cases and counterfactual wrong-repository controls.
• Agent Retrieval Bench contains 427 samples across 25 repositories: 345 positive examples, 50 natural no-gold examples, and 32 counterfactual controls.

Results & evidence

• arXiv:2607.24882v1 Announce Type: cross Abstract: Modern coding agents are usually evaluated by whether they eventually produce a correct patch, but patch generation depends on an earlier context-acquisition stage: finding the repository files needed for th...
• Agent Retrieval Bench contains 427 samples across 25 repositories: 345 positive examples, 50 natural no-gold examples, and 32 counterfactual controls.
• The corpus includes 308 base-commit snapshots, 392,000 files, and 7.9 million chunks.

Limitations / unknowns

• Generalization outside curated tasks is still unclear.

Next-step validation checks

• Reproduce one claim with a public baseline and fixed evaluation settings.
• Check robustness on out-of-distribution or long-context cases.
• Track whether independent teams report matching results.

Context

You seed baits, artifacts crafted to look exactly like the high-value, security-relevant assets such an agent is hunting for, staged in the places it will look and wrapped in the context that sells them: a plausible filename, a companion ciphertext blob, a...

What's new

by Cracken; Core team: Dario Pasquini and Michal Bazyli Blacksea is an active honeypot and canary-bait control system built to detect and drown LLM-driven attackers: autonomous AI agents and LLM-assisted operators that scan and exploit systems.

Key details

• Blacksea doesn't stop at watching LLM attacks.
• It exploits flaws in the attacker's LLM judgment to gain arbitrary code execution on their machines, collect intel passive defenses can't reach, and make sure they don't come back.
• An LLM-driven attacker works an engagement by reasoning toward the assets that move it forward: credentials to reuse, a decryptor for an encrypted blob, a key-derivation tool, a config unpacker, a token minter, an internal API client.
• Blacksea turns that reasoning into a trap.

Results & evidence

• No hard numbers surfaced in the source text; treat claims as directional until benchmarks appear.

Limitations / unknowns

• Generalization outside curated tasks is still unclear.

Next-step validation checks

• Reproduce one claim with a public baseline and fixed evaluation settings.
• Check robustness on out-of-distribution or long-context cases.
• Track whether independent teams report matching results.

Context

arXiv:2607.24882v1 Announce Type: cross Abstract: Modern coding agents are usually evaluated by whether they eventually produce a correct patch, but patch generation depends on an earlier context-acquisition stage: finding the repository files needed for th...

What's new

arXiv:2607.24882v1 Announce Type: cross Abstract: Modern coding agents are usually evaluated by whether they eventually produce a correct patch, but patch generation depends on an earlier context-acquisition stage: finding the repository files needed for th...

Key details

• We introduce Agent Retrieval Bench, a file-level benchmark for this upstream retrieval problem.
• Samples are built from real coding-workflow signals and evaluated against frozen base-commit repositories, with relevance defined by what an agent needs next rather than direct query-file semantic similarity.
• The benchmark covers four positive-retrieval tasks: code2test, comment2context, trace2code, and edit2ripple; a fifth subset evaluates selective retrieval using natural evidence-backed no-gold cases and counterfactual wrong-repository controls.
• Agent Retrieval Bench contains 427 samples across 25 repositories: 345 positive examples, 50 natural no-gold examples, and 32 counterfactual controls.

Results & evidence

• arXiv:2607.24882v1 Announce Type: cross Abstract: Modern coding agents are usually evaluated by whether they eventually produce a correct patch, but patch generation depends on an earlier context-acquisition stage: finding the repository files needed for th...
• Agent Retrieval Bench contains 427 samples across 25 repositories: 345 positive examples, 50 natural no-gold examples, and 32 counterfactual controls.
• The corpus includes 308 base-commit snapshots, 392,000 files, and 7.9 million chunks.

Limitations / unknowns

• Generalization outside curated tasks is still unclear.

Next-step validation checks

• Reproduce one claim with a public baseline and fixed evaluation settings.
• Check robustness on out-of-distribution or long-context cases.
• Track whether independent teams report matching results.

Context

We show the mismatch to be an evaluation problem in the costume of a training one at a fixed recall, prior correction and calibration cannot move precision, and answer it with a prior-matched reporting method based on three figures: balanced-test, operation...

What's new

We show the mismatch to be an evaluation problem in the costume of a training one at a fixed recall, prior correction and calibration cannot move precision, and answer it with a prior-matched reporting method based on three figures: balanced-test, operation...

Key details

• Because attention is the cost of error, precision leads.
• Its classifier was trained and reported at a one-to-one class balance, fixed before the operational rate could be known.
• That rate has since emerged at roughly one scene in twenty, and a balanced-test score badly overstates the precision a validator meets.
• A model that scores 0.794 balanced-test precision scores 0.192 in real operation: the gap is a systematic artefact of reporting at the wrong prior, invisible to the metric most work quotes.

Results & evidence

• arXiv:2607.07146v2 Announce Type: replace Abstract: The Internal Waves Service screens the Sentinel-1 Wave-mode archive for internal solitary waves, routing detections to experts whose adjudication time is the resource the effort exists to conserve.
• A model that scores 0.794 balanced-test precision scores 0.192 in real operation: the gap is a systematic artefact of reporting at the wrong prior, invisible to the metric most work quotes.
• Holding recall at a floor of 0.80 and certifying against a sealed, single-read lockbox, the promoted model reports 0.927 precision at the operational prior; an out-of-time check confirms discrimination transfers to unseen periods while a fixed operating poi...

Limitations / unknowns

• Generalization outside curated tasks is still unclear.

Next-step validation checks

• Reproduce one claim with a public baseline and fixed evaluation settings.
• Check robustness on out-of-distribution or long-context cases.
• Track whether independent teams report matching results.

Context

Even when provided with the same contextual information, an LLM may generate a correct patch for one bug but fail on another closely related bug.

What's new

In this paper, we present the first empirical study of attention patterns in LLM-based program repair, providing interpretable insights into how models process bug reports and where their attention is concentrated during repair.

Key details

• Even when provided with the same contextual information, an LLM may generate a correct patch for one bug but fail on another closely related bug.
• Why this happens remains poorly understood, and it is unclear how LLMs prioritize the diverse information in bug reports and whether model attention affects repair success.
• In this paper, we present the first empirical study of attention patterns in LLM-based program repair, providing interpretable insights into how models process bug reports and where their attention is concentrated during repair.
• We analyze 319 real-world Python and Java bugs from SWE-bench Verified and Multi-SWE-bench to study (RQ1) how model attention is distributed across bug report sections, (RQ2) how attention patterns within each section differ between successful and unsuccess...

Results & evidence

• arXiv:2607.25873v1 Announce Type: cross Abstract: Large Language Model (LLM)-based Automated Program Repair systems are advancing rapidly, yet their performance remains inconsistent.
• We analyze 319 real-world Python and Java bugs from SWE-bench Verified and Multi-SWE-bench to study (RQ1) how model attention is distributed across bug report sections, (RQ2) how attention patterns within each section differ between successful and unsuccess...
• Computer Science > Software Engineering [Submitted on 28 Jul 2026] Title:How Do LLMs Read Bug Reports?

Limitations / unknowns

• Why this happens remains poorly understood, and it is unclear how LLMs prioritize the diverse information in bug reports and whether model attention affects repair success.
• We find that successful repairs are characterized by diffused attention across multiple diagnostic components such as bug descriptions, stacktraces, and test cases, while failures often exhibit over-localized attention toward metadata such as version inform...
• Our results provide the first empirical evidence that attention misallocation is a key factor in LLM-based APR failures, and offer actionable insights for designing more interpretable and reliable future APR systems.

Next-step validation checks

• Reproduce one claim with a public baseline and fixed evaluation settings.
• Check robustness on out-of-distribution or long-context cases.
• Track whether independent teams report matching results.

Context

Compress tool outputs, logs, files, and RAG chunks before they reach the LLM.

What's new

Compress tool outputs, logs, files, and RAG chunks before they reach the LLM.

Key details

• 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers.
• ██╗ ██╗███████╗ █████╗ ██████╗ ██████╗ ██████╗ ██████╗ ███╗ ███╗ ██║ ██║██╔════╝██╔══██╗██╔══██╗██╔══██╗██╔═══██╗██╔═══██╗████╗ ████║ ███████║█████╗ ███████║██║ ██║██████╔╝██║ ██║██║ ██║██╔████╔██║ ██╔══██║██╔══╝ ██╔══██║██║ ██║██╔══██╗██║ ██║██║ ██║██║╚██╔...
• Headroom compresses everything your AI agent reads — tool outputs, logs, RAG chunks, files, and conversation history — before it reaches the LLM.
• Same answers, fraction of the tokens.

Results & evidence

• 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers.
• Live: 10,144 → 1,260 tokens — same FATAL found.
• - Library — compress(messages)in Python or TypeScript, inline in any app - Proxy — headroom proxy --port 8787, zero code changes, any language - Agent wrap — headroom wrap claude|codex|grok|copilot|cursor|aider|opencode|cline|continue|goose|openhands|opencl...

Limitations / unknowns

• Generalization outside curated tasks is still unclear.

Next-step validation checks

• Reproduce one claim with a public baseline and fixed evaluation settings.
• Check robustness on out-of-distribution or long-context cases.
• Track whether independent teams report matching results.

Context

AI agent skill that researches any topic across Reddit, X, YouTube, HN, Polymarket, and the web - then synthesizes a grounded summary An AI agent-led search engine scored by upvotes, likes, and real money - not editors.

What's new

AI agent skill that researches any topic across Reddit, X, YouTube, HN, Polymarket, and the web - then synthesizes a grounded summary An AI agent-led search engine scored by upvotes, likes, and real money - not editors.

Key details

• This README tracks the current v3 pipeline.
• The runtime skill spec lives in skills/last30days/SKILL.md, which is the source of truth for the latest command and setup behavior.
• Claude Code (recommended — auto-updates via marketplace): /plugin marketplace add mvanhorn/last30days-skill /plugin install last30days Codex, Cursor, Copilot, Gemini CLI, or any of 50+ Agent Skills hosts: npx skills add mvanhorn/last30days-skill -g (-g inst...
• Drop it to scope per-project.) More install options (claude.ai web, OpenClaw, manual) in the Install section below.

Results & evidence

• Claude Code (recommended — auto-updates via marketplace): /plugin marketplace add mvanhorn/last30days-skill /plugin install last30days Codex, Cursor, Copilot, Gemini CLI, or any of 50+ Agent Skills hosts: npx skills add mvanhorn/last30days-skill -g (-g inst...
• Run it once and the setup wizard unlocks X, YouTube, TikTok, arXiv, Techmeme, and more in 30 seconds.

Limitations / unknowns

• Generalization outside curated tasks is still unclear.

Next-step validation checks

• Reproduce one claim with a public baseline and fixed evaluation settings.
• Check robustness on out-of-distribution or long-context cases.
• Track whether independent teams report matching results.

Context

arXiv:2607.24868v1 Announce Type: new Abstract: This report studies noise-shaped one-bit coefficients in normalized discrete polynomial Fourier extension.

What's new

arXiv:2607.24868v1 Announce Type: new Abstract: This report studies noise-shaped one-bit coefficients in normalized discrete polynomial Fourier extension.

Key details

• For first-order Sigma-Delta quantization, the error is written as $e_k=u_k-q_k=\Delta v_k$ with a uniformly bounded state.
• Discrete summation by parts then yields variation estimates for complex weights and an $O(N^{-1})$ approximation rate on compact parameter sets.
• For the parabolic phase $\phi_{x,t}(\xi)=x\xi+t\xi^2$, the bound is expressed through $J(x,t)=\int_0^1 |x+2t\xi|d\xi$, and the uniform $N^{-1}$ rate is shown to be sharp over the admissible input class.
• Higher-order finite-record identities are derived with all endpoint traces retained.

Results & evidence

• arXiv:2607.24868v1 Announce Type: new Abstract: This report studies noise-shaped one-bit coefficients in normalized discrete polynomial Fourier extension.
• Discrete summation by parts then yields variation estimates for complex weights and an $O(N^{-1})$ approximation rate on compact parameter sets.
• For the parabolic phase $\phi_{x,t}(\xi)=x\xi+t\xi^2$, the bound is expressed through $J(x,t)=\int_0^1 |x+2t\xi|d\xi$, and the uniform $N^{-1}$ rate is shown to be sharp over the admissible input class.

Limitations / unknowns

• Generalization outside curated tasks is still unclear.

Next-step validation checks

• Reproduce one claim with a public baseline and fixed evaluation settings.
• Check robustness on out-of-distribution or long-context cases.
• Track whether independent teams report matching results.

Context

Enprompta – Prompt Registry, LLM Evals, and Observability for Production AI Apps

What's new

Enprompta – Prompt Registry, LLM Evals, and Observability for Production AI Apps

Key details

• Enprompta – Prompt Registry, LLM Evals, and Observability for Production AI Apps

Results & evidence

• No hard numbers surfaced in the source text; treat claims as directional until benchmarks appear.

Limitations / unknowns

• Generalization outside curated tasks is still unclear.

Next-step validation checks

• Reproduce one claim with a public baseline and fixed evaluation settings.
• Check robustness on out-of-distribution or long-context cases.
• Track whether independent teams report matching results.

Context

Show HN: ButterClaw – AI agent runtime security, SIGKILL on breach, no cloud

What's new

Show HN: ButterClaw – AI agent runtime security, SIGKILL on breach, no cloud

Key details

• Show HN: ButterClaw – AI agent runtime security, SIGKILL on breach, no cloud

Results & evidence

• No hard numbers surfaced in the source text; treat claims as directional until benchmarks appear.

Limitations / unknowns

• Generalization outside curated tasks is still unclear.

Next-step validation checks

• Reproduce one claim with a public baseline and fixed evaluation settings.
• Check robustness on out-of-distribution or long-context cases.
• Track whether independent teams report matching results.

Context

I was getting tired of all my sites having the same generic fonts and having AI generated writing because it's cumbersome to experiment.

You can ask an agent to give you some font ideas but they link out to a specimen where you imagine what it might...

What's new

I was getting tired of all my sites having the same generic fonts and having AI generated writing because it's cumbersome to experiment.

You can ask an agent to give you some font ideas but they link out to a specimen where you imagine what it might...

Key details

• You ask the agent to wire it up and it looks different than what you thought.

  I just wanted to be able to play with fonts in the project itself, as well as the ability to edit the text where I saw it, on my live project via local dev server without needin...

• And then the one I pick is exactly what I get in the live environment.

  Font Lab lets you see several font options right in your site.

• If the agent can access your local computer (ex: cursor), it can install the Font Lab panel, curate options from any Google font, and then let you swap them in real time on your localhost dev server.
• If you're using web or a non-supported framework, it can generate screenshots with the fonts on your site so you can still really see what it would look like for your project.

  You can also double click on any text on your site to edit in-place.

Results & evidence

• No hard numbers surfaced in the source text; treat claims as directional until benchmarks appear.

Limitations / unknowns

• Generalization outside curated tasks is still unclear.

Next-step validation checks

• Reproduce one claim with a public baseline and fixed evaluation settings.
• Check robustness on out-of-distribution or long-context cases.
• Track whether independent teams report matching results.