Software Engineering News
An unofficial font recreating the classic Commodore 64 keyboard's printed key legends, including PETSCII graphics, has been released with download and preview tools, from hackernews.
A new font called C64 Keyboard reconstructs the legends printed on classic Commodore 64 keycaps, based on photographs of the keyboard but with cleaned-up shapes and consistent proportions. It's an independent, unofficial project — not exact photo traces, but inferred outlines. Version 1.107 is current, and anyone with an older install should replace it.
The package covers uppercase letters, numbers, punctuation, arrows, the £ sign, selected accented letters including Hungarian characters, function-key labels f1–f12, and all 63 PETSCII graphic front legends. Lowercase input displays as uppercase, matching how the original keys were printed. The PETSCII symbols are arranged by key in the preview, with left legends corresponding to Commodore + key and right ones to Shift + key in uppercase/graphics mode.
To use it, install the TTF or OTF and select the font in your application, or open preview.html from the complete package to test it. A type tester handles the special character codes for PETSCII symbols, though copied symbols only display correctly with this font installed. The font is meant to evoke the printed key legends, not the C64's bitmap screen font.
Editable outlines are available in a glyphs/ directory with build tools in scripts/, and the identities and key assignments draw on the Ultimate Commodore 64 Reference. Original photographs aren't included in the release.
Source: news.ycombinator.com, github.com
An AI lab released a batch of new mathematical results from an internal frontier model, with proof formalizations and transparency details, on GitHub from hackernews
The lab published a set of new mathematical results produced by an internal frontier model, released via a GitHub repository with protocols for revisions and citations. They consulted the independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study, drawing on its public recommendations to shape the release, and say they're still looking at other community-hosted venues that meet the committee's guidelines.
The repository includes formalizations of many proofs in Lean, the machine-checkable proof language, with more to be added over time. On the transparency side, they published 10 summaries of the model's reasoning, compute estimates — the average result took roughly the equivalent of three hours of ChatGPT Pro thinking — and statistics on how many problems the model attempted.
Beyond the papers themselves, the lab is funding workshops, conferences, and special programs built around understanding AI-produced results, and says it's working toward a responsible release of the model behind them. Commitments for future releases include better mathematical exposition and citations, plus continued evaluation of frontier models on mathematics and other sciences.
Source: news.ycombinator.com, openai.com
Strands Labs released Strands Decider 2B, an open-source 2-billion-parameter "decision model" for fast, local classification tasks, from hackernews
Strands Labs shipped Strands Decider 2B, a small open-source model built for a new class of "decision models" (or "system one" models) — a category that's gotten attention since TypeSafe AI launched Jev earlier this month. Unlike LLMs, these models don't generate arbitrary text; they pick between given options or assign numerical scores, and they attach a reliability score to each decision. The tradeoff: much faster and better-calibrated at a given size, but useless for chat, coding, or summarization, and worse than reasoning models on complex problems.
Architecturally, it's a Qwen3.5-2B torso with the LM head removed and replaced by a small (~1M parameter) pointer head that scores each option's hidden state against the answer position, fine-tuned with a rank-16 LoRA adapter. It's the 19th iteration of the architecture; an earlier "slot head" design performed notably worse. On JevBench's public set it ranks 3rd of 33 in the 2B class on accuracy (1st of 30 excluding just-over-2B models), with median decision latency around 115ms on an RTX 3090 and ~153ms on an M3 MacBook.
The team sized it at 2B deliberately — small enough to train and run on local hardware, big enough to handle all the "easy" tasks on JevBench correctly. Intended uses include model routing, tool selection, guardrails, evaluations, and hybrid agents where the decider handles rote decisions while an LLM handles the hard ones. Weights, training data, and scripts are on GitHub and Hugging Face, with a pip-installable CLI to try it immediately.
Source: news.ycombinator.com, strandsagents.com
OpenAI launched a public beta of a Decisions API that returns typed answers—probabilities, categorical choices, or scores—about text or images roughly 10x faster than the Responses API, from hackernews
The new endpoint, POST /v1/decisions, currently runs only on a model called gpt-6-luna and takes three inputs: a model, shared input evidence (text or user messages mixing text and base64 images), and a questions array. Each question gets a type: predicate (returns a 0–1 probability a condition is true), choice (picks one of your supplied categories, with per-option probabilities and a confidence field), or score (rates against ordered levels via a probability-weighted average, so scores can land between levels).
The intended use is classification, routing, and prioritization inside applications rather than content generation. Examples include flagging damaged product photos for review, routing a complaint to billing versus shipping, and scoring bug severity. Multiple independent questions can run against shared input in one request; dependent decisions require separate calls. Images must be inline base64 data URLs—hosted URLs and file IDs aren't supported. SDK support requires recent versions across Python, JavaScript, Go, Ruby, and Java.
OpenAI positions this alongside existing tools: use Structured Outputs when you need generated JSON or explanations, function calling for tool requests, and Decisions when you need one of these answer types. Guidance emphasizes writing questions around observable criteria, including fallback options like "other," and setting thresholds with labeled examples based on your false-positive and false-negative costs. GA is expected in the coming weeks; the article was cut off mid-pricing, but input costs $0.10 per some unit (likely per million tokens).
Source: news.ycombinator.com, developers.openai.com
Google released EmbeddingGemma 2, an open Apache 2.0 model that maps text, images, audio, video, and code into one embedding space for on-device search and RAG, from hackernews.
EmbeddingGemma 2 is Google's follow-up to last year's EmbeddingGemma, which saw over 20 million downloads. The new model, built on the Gemma 4 architecture at 740M parameters, natively handles code, images, video, and audio alongside text — tasks like finding a video clip from a voice memo or searching hours of audio by text query run locally.
The design is modular: text-only workloads need as little as 270M parameters, with optional vision (170M) and audio (300M) encoders. Matryoshka Representation Learning lets developers truncate embeddings from 768 down to 128 dimensions for up to 6x storage savings. Quantized, it runs on a Pixel 11 Pro in roughly 191MB of RAM for text or 567MB fully multimodal.
Performance-wise, it leads sub-1B multimodal embedders on MTEB Code and MAEB benchmarks, with a 9.92-point code improvement over its predecessor and an 8K token context (4x larger) supporting 5.5 minutes of audio or 58 video frames per input. It shares a tokenizer and audio encoder with Gemma 4, so the two can run together in one pipeline with a smaller combined footprint.
Weights are on Hugging Face and Kaggle, with support from MediaPipe, LiteRT, transformers.js, Ollama, vLLM, and fine-tuning guidance from Unsloth.
Source: news.ycombinator.com, blog.google
A Pi-hole-style DNS ad-blocker runs on a $2 ESP32-C3 by storing blocklist domains as 40-bit hashes in flash instead of domain strings in RAM, from hackernews.
Most hobbyist DNS sinkholes for the ESP32 need PSRAM because they load the blocklist as domain strings into memory. This project sidesteps that by storing sorted 5-byte FNV-1a hashes in flash and binary-searching them: 141k domains fit in 0.67 MB of flash, lookups take ~10 ms, and only ~50 KB of RAM is used. Collisions are effectively zero at 141k domains and about one at 537k.
The trick isn't limited to the cheap chip — hashes in flash beat strings in PSRAM on bigger boards too (roughly 2.7M domains on a 16 MB ESP32-S3 vs 466k for strings). Supported hardware includes any ESP32-C3 with 4 MB flash plus the classic ESP32 DevKit, and there's a printable enclosure with antenna and venting notes.
It's a fairly complete package: captive-portal WiFi setup, a dashboard with blocklist and firmware OTA updates, a weekly auto-rebuilt blocklist hosted on GitHub Releases, and support for hosts files, plain domain lists, and AdGuard-style rules (regex and cosmetic rules are skipped). Firmware OTA costs flash space though — dual app slots cap the blocklist at ~250k domains, so the 537k list requires dropping OTA.
Security was tightened after an earlier version left state-changing endpoints open to anyone on the LAN; Basic Auth now gates them, and a stored-XSS path in custom domain names was closed. But it's plain HTTP with base64 credentials — the author is upfront that this guards against casual LAN abuse, not an on-path attacker, and notes an unresolved CSRF concern where browsers auto-attach cached Basic Auth to cross-site requests.
Source: news.ycombinator.com, github.com
Pi's new Codemode lets LLM agents orchestrate harness-side operations via JavaScript, complementing bash for tasks like image generation and classification, from hackernews.
The author previously argued against custom tools and MCP servers, pushing scripts instead. But Pi 1.0 now ships MCP support through Codemode—a JavaScript interface that runs in the harness's sandbox (QuickJS in WASM, no network/filesystem/timers) and issues tool calls without polluting the LLM's context. Bash remains for execution-environment tasks, but Codemode handles harness-native operations like reading images or spawning subagents, where the model's "brain" and the sandboxed "hands" are separate systems.
Codemode's key advantages: larger tool outputs get structured handling rather than trailing 2000-line truncation, concurrent operations are expressible, and state can be stashed in the transcript between invocations. The agent typically probes 5-10 items first, then writes a script to process the rest. Internal Pi APIs—image generation and classifier models like Jev—are exposed only via Codemode, not regular tools, so agents can generate images or mass-classify GitHub issues (e.g., sentiment analysis of 100 issues) directly from JavaScript.
Source: news.ycombinator.com, lucumr.pocoo.org
nickvasilescu/bops is trending in typescript (116 stars) from GitHub
Bops: a team of AI bots that run your business ops, each with its own computer, email and phone number. bops.bot
Source: github.com
accomplish999/position-sizer is trending in typescript (63 stars) from GitHub
Position size calculators for Perps and DeFi.
Source: github.com
sisyphuslabs/omo-dori-mode-experimental is trending in typescript (49 stars) from GitHub
Dori mode: an always-on messenger agent that launches, tracks and closes coding-agent sessions in herdr (skill + bun CLI). Experimental.
Source: github.com
nullmoth/nvidia-macos-driver is trending in rust (73 stars) from GitHub
Metal driver for NVIDIA GeForce RTX cards on macOS 15 Sequoia (Intel / OpenCore). Free, source included.
Source: github.com
storytold/craft-fonts is trending in rust (47 stars) from GitHub
Fonts for the Crafting Apps
Source: github.com
storytold/wordcraft is trending in rust (38 stars) from GitHub
An open-source, clean-room reimplementation of Microsoft Word in pure Rust
Source: github.com
saidi-walid-architecte/saidi-walid-architecte.github.io is trending in python (100 stars) from GitHub
Site de Saidi Walid, architecte agréé et expert judiciaire à Batna
Source: github.com
jinrunsen/claude-my-privacy is trending in python (56 stars) from GitHub
MyPrivacy: local name and company pseudonymization for Claude Code, with restoration for tools and assistant reply display.
Source: github.com
GrimClawBot/Pixel-Companion is trending in python (46 stars) from GitHub
Native macOS notch companion for agent runtimes, with an optional Pixel HQ/Paperclip connector.
Source: github.com
AI News
Jump Trading is using OpenAI's models to scale its quantitative research workflows, combining longer-running AI tasks across multiple data sources with human oversight, from OpenAI Blog.
The article describes how Jump Trading, the Chicago-based trading firm, has turned to OpenAI's models — including ChatGPT — to expand what its quantitative research teams can cover. Rather than replacing researchers, the setup runs AI workflows that pull together several data sources at once and operate over longer stretches of time than a typical chat session.
Human review remains a core part of the loop. The researchers treat the models as a way to widen the funnel of hypotheses and analyses worth examining, with people still checking the outputs before anything informs trading decisions.
The piece is framed as a case study in how firms outside the software industry are adapting AI tooling for specialized, high-stakes work — a signal that quant shops see off-the-shelf models as useful for research productivity, not just demos.
Source: openai.com
OpenAI published new results on open math problems from an internal frontier model, along with Lean proof formalizations and research details on GitHub, from OpenAI Blog
OpenAI released results showing an internal frontier model making progress on open problems in mathematics. Alongside the results, the company published formalizations of the proofs in Lean, the proof assistant used by formal mathematics researchers, so the work can be checked and built on rather than taken on trust.
The supporting material — proof formalizations and research details — went up on GitHub, which puts the artifacts in reach of the math and formalization communities directly. That's the part worth watching: machine-generated proofs of open problems are only as credible as their verifiable form, and shipping Lean code is how that credibility gets established.
It's also a notable publishing choice for a company that typically keeps frontier model work internal. Whether this becomes a pattern or a one-off disclosure will say something about how OpenAI handles research communication going forward.
Source: openai.com
OpenAI announced a partnership with Ironclad to train and evaluate AI agents on complex contracting workflows, advancing computer use for professional work, from OpenAI Blog.
OpenAI is working with Ironclad, a contract lifecycle management company, to develop AI agents that can handle real contracting work. The piece centers on using Ironclad's domain — legal contracts, with their multi-step review, drafting, and approval processes — as a testbed for computer use capabilities.
The rationale is that contracting workflows are a good proving ground for agents: they involve professional judgment, long sequences of actions across software interfaces, and real consequences for errors. Training and evaluating on this kind of work pushes agents beyond simple tasks toward what actual knowledge workers do.
What's described here is infrastructure for capability-building — using a real company's workflows to benchmark and refine how agents operate computers — rather than a product launch. Readers watching the agent space will want more detail on evaluation results, which this summary-level post doesn't yet provide.
Source: openai.com
Atlassian and OpenAI are broadening their partnership to connect frontier AI models with enterprise knowledge so teams can plan, build, and deliver work, from OpenAI Blog.
Atlassian and OpenAI announced an expanded partnership aimed at putting OpenAI's frontier models to work on enterprise knowledge. The stated goal is helping teams plan, build, and deliver work, rather than just search or summarize it.
For Atlassian, this deepens its AI positioning across tools like Jira and Confluence, where customers store much of their day-to-day project and documentation data. For OpenAI, it's another enterprise distribution channel at a time when rivals like Anthropic and Google are courting the same corporate buyers.
The announcement is short on specifics—no product names, timelines, or integration details were given. Expect concrete rollout news, if any, to come from Atlassian's side in follow-up announcements.
Source: openai.com
Falcon-Emirati: When an LLM Learns the Dialect, the Culture, and the Nuance from Hugging Face Blog
Source: huggingface.co
The Agent Said It Was Done. The Database Disagreed. from Hugging Face Blog
Source: huggingface.co
Open-sourcing AstaBrief, the fast report-generation model in Asta from Hugging Face Blog
Source: huggingface.co
AutoSynthData: Generating Training Data for Enterprise Agents from Hugging Face Blog
Source: huggingface.co
Former Ramp engineers have raised $20 million for Melius, a startup that pivoted from ad spend optimization to building tools that generate creative assets and campaigns, from TechCrunch AI.
A team of ex-Ramp engineers has secured $20 million in funding for Melius, their new startup. The raise comes after the founders scrapped their first product, which was aimed at helping marketers manage and optimize ad spend.
The company has now shifted its focus to a different part of the marketing stack: tools that generate the creative assets and campaigns themselves, rather than managing the money behind them. The pivot puts Melius in the increasingly crowded space of AI-powered creative generation for advertisers.
Source: techcrunch.com
Musubi released open weights for PolicyLM-1.7B, a lightweight AI decision model built for real-time content moderation, from TechCrunch AI.
Musubi announced PolicyLM-1.7B on Tuesday — a small decision model aimed specifically at real-time content moderation, and notably released with open weights rather than kept behind an API. The 1.7B parameter size is the practical point here: small enough to run cheaply and fast, which is what real-time moderation actually requires.
Open weights are the part worth watching. Moderation tooling has largely been the province of platform-internal systems, so a model anyone can inspect or self-host shifts where that capability lives.
It's early — the announcement covers the release itself, and how PolicyLM-1.7B performs on real moderation workloads is still an open question.
Source: techcrunch.com
Lambda is raising up to $4 billion at a $14.5 billion pre-money valuation, led by Coatue and Blackstone, as it prepares for a 2027 IPO from TechCrunch AI.
Lambda, the AI computing company with Nvidia on its cap table, is lining up a round of up to $4 billion. The raise values the company at $14.5 billion pre-money, with Coatue and Blackstone leading the deal.
The funding comes as Lambda lines up a public listing targeted for 2027, making this likely a final major private round before it faces public-market scrutiny.
Lambda sells GPU cloud capacity and hardware to companies training AI models, putting it in direct competition with CoreWeave, which went public earlier this year. Coatue and Blackstone backing it signals continued investor appetite for AI infrastructure plays, even at steep valuations.
Source: techcrunch.com
AI agents that promise to shop and book for users are being blocked by websites' anti-bot defenses, and a new standard aims to fix it, from TechCrunch AI.
AI agents that can shop, book flights, and make reservations on your behalf keep hitting a wall: websites are deliberately blocking them. Anti-bot defenses built to stop scrapers and credential-stuffing attacks can't tell a helpful personal agent from malicious automation, so they shut it down.
That leaves consumers stuck in the middle — they've delegated the task to an agent, but the agent can't complete it because the site refuses it access. The friction is growing as agent tools reach mainstream users and hit more sites designed to keep automated traffic out.
A new standard is being proposed to solve this, giving websites a way to recognize and permit legitimate AI agents rather than blanket-blocking them. Whether sites adopt it is the open question: letting agents in means rethinking defenses that many businesses rely on.
Source: techcrunch.com
Common Sense Media called OpenAI's ChatGPT for Teens an "unacceptable risk," citing weak safety guardrails, from The Verge AI.
Common Sense Media, the nonprofit that reviews apps and services with a focus on youth safety, published an assessment of ChatGPT for Teens, the teen-specific version OpenAI launched in August with guardrails meant to help students learn safely.
The group's findings were blunt: the protections "fall short of what the company promised." ChatGPT fails to alert parents when it should, doesn't provide the right help in crisis situations, and still completes kids' homework despite the teen framing.
The "unacceptable risk" label is the takeaway from the nonprofit's press release, putting pressure on OpenAI to close the gap between its stated safeguards and how the product actually behaves for younger users.
Source: theverge.com
OpenAI says an unreleased frontier model solved hundreds of open math problems across 722 manuscripts, from The Verge AI
OpenAI has released a batch of 722 manuscripts — organized into 372 related result families — containing solutions to long-standing mathematics problems, all produced by a frontier model the company has not yet made public. According to AGMAI, an independent advisory group of elite mathematicians formed to handle communication of the results, the batch includes solutions to "hundreds" of open questions.
The release had been anticipated for weeks and extends a streak of mathematical breakthroughs that has drawn attention — and unease — from parts of the math community.
Alongside the excitement, the work is raising questions about research ethics and academic conduct, which is presumably part of why OpenAI routed the announcement through AGMAI rather than publishing the results outright.
Source: theverge.com
Tech companies are redefining what counts as "recording" to sell AI hardware, as Apple's latest smart home plans show, from The Verge AI.
For decades the rule was simple: a device with a mic or camera either captured and stored data or it didn't. That data—saved locally or shipped to the cloud for processing—was a recording. Mark Gurman's Bloomberg report on Apple's smart home hardware last week pushed against that assumption, and The Verge argues companies are now blurring the line on purpose.
The pitch from AI device makers is that always-listening sensors aren't "recording" if the audio isn't saved—just processed, transcribed, or discarded after use. That framing lets hardware sit in bedrooms and living rooms without technically being a recorder, while still capturing everything needed to make the AI features work.
The article's argument is that this is a definitional dodge, not a real distinction. Whether data persists for thirty seconds or thirty years may matter less than the fact that it was captured, transmitted, and could be retained. Consumers have long relied on a bright-line rule—recording on or off—and AI hardware is quietly erasing it while marketing does the explaining.
Source: theverge.com
Starting October 9th, free Gemini users will be limited to Flash Lite, with Flash requiring a $4.99/month Google AI Plus subscription and Pro reserved for higher tiers, from The Verge AI
Beginning October 9th, Google is cutting back what free Gemini users can access. The free plan will be restricted to Flash Lite only, whereas free users can currently pick between Flash Lite, Flash, and Pro.
Getting the standard Flash model back will require a $4.99/month Google AI Plus subscription — but that tier is itself getting weaker, since it will soon drop Gemini Pro from its included features. Google says AI Plus subscribers will get an email notifying them of when they lose Pro access.
The top-tier Gemini Pro model and the "Deep Think" advanced reasoning option will be gated behind Google AI Pro or Ultra subscriptions, the pricier plans in Google's lineup.
Source: theverge.com
Mistral announced a public preview of Mistral Large 4, a 1-trillion-parameter multimodal model with open weights coming at month's end, from hackernews.
Mistral launched a preview API for ML4 (nicknamed "le Chonk"), its largest model yet — 1T parameters total, 49B active, natively multimodal. Weights are promised by end of the month after a red-teaming period with cybersecurity partners, some of whom get reduced moderation and expanded cyber capabilities in the meantime.
The performance claims lean heavily on cybersecurity and agentic work. It scores 82% on one Artificial Analysis Cyber test where several closed models (Claude Opus 5.5, GPT-6 Astra) score near zero — not because they can't, but because they refuse to reproduce and patch real vulnerabilities. Mistral frames this as a defensible advantage for incident response, and the open weights let security teams run the model on-prem without losing access mid-incident. On coding it posts 61.7% on DeepSWE and 49.8% on the Coding Agent Index, ahead of DeepSeek V4 Pro and Qwen3.8 Max, and ranked second behind Claude Opus 5 in blind human evals.
Everything was trained from scratch on 3,800 NVIDIA Grace Blackwell GPUs in Mistral's own European datacenters, with a fully sovereign European deployment planned. Training data covered 160+ languages, including every official EU language. Architecture details, more benchmarks, and post-training methodology are promised before the weight release.
Source: news.ycombinator.com, mistral.ai
Politics
Coverage by lean: ● covered it · ○ not in this digest's sources
| Story | Left | Center | Right |
|---|
| Texas Senate race: Paxton vs. Talarico draws national attention | ● | ● | ● |
| Supreme Court hears oral arguments in Suncor v. Boulder climate lawsuit | ● | ○ | ● |
| First Maine Senate debate between Susan Collins and Troy Jackson | ● | ● | ● |
| Trump Does Not Plan to Reimburse Taxpayers for Promotional Ads | ● | ○ | ○ |
| Ex-CIA officer David Rush pleads guilty to defrauding government of millions | ● | ● | ○ |
| Trump suggests Iran could 'take out' Los Angeles and San Diego at rally | ● | ● | ● |
| An Influential Voice in the Climate Debate Has Died at 66 | ● | ○ | ○ |
| Olympian wrongfully indicted in Reflecting Pool case seeks inspector general probes | ● | ● | ○ |
| Trump announces $6.6B Anduril shipyard in Baltimore for submarine components | ● | ○ | ● |
| GOP Super PAC pours $5 million into South Carolina Senate race, sources say | ● | ○ | ○ |
| Fort Hood shooter Nidal Hasan scheduled for December 3 execution by firing squad | ● | ● | ○ |
| ABC seeks court order blocking FCC's early review of broadcast licenses | ● | ● | ○ |
| Trump administration sends $90 Medicare and $500 ACA refund checks | ● | ● | ○ |
| The surprising, cut-throat race for Los Angeles mayor | ● | ○ | ● |
| Lawsuit alleging drugged gang rape at Cornell fraternity prompts legal and political debate | ○ | ● | ● |
| AOC fundraiser remarks frame Abdul El-Sayed's Michigan Senate race | ○ | ● | ● |
Trump Officials Ramp Up Investigations Into Voter Registration Groups from NYT vs Justice Dept. Turns to Partisans and Conspiracists to Fill Depleted Ranks from NYT
The Trump administration is investigating nonprofits that it says may be recruiting noncitizens for voter registration, documents show, even though noncitizen voting in the United States is rare.
Source: NYT (lean-left, nytimes.com), NYT (lean-left, nytimes.com)
Texas Senate race: Paxton vs. Talarico draws national attention
Left · NYT † — Focuses on Angela Paxton supporting her estranged husband's Senate campaign despite divorce and adultery accusations, noting her own political ambitions.
Center · The Hill † — Frames the race around Trump's Texas visit and GOP worry as Paxton trails Talarico amid midterm forecasts shifting toward Democrats.
Right · Fox News — Spotlights an unearthed Talarico clip on Texas schools, framing his education record and progressive positions as campaign liabilities. Left out: The Hill and NYT noted Trump's Texas visit and Angela Paxton's role; Fox News did not mention either.
† Based on the headline and RSS summary only; the full article wasn't available.
Source: NYT (lean-left, nytimes.com), Fox News (lean-right, foxnews.com), The Hill (center, thehill.com)
Supreme Court hears oral arguments in Suncor v. Boulder climate lawsuit
Left · NYT † — Frames the case through the environmental stakes, highlighting the Air Force's long-running bomb disposal practice and its impact on a Guam beach habitat.
Left · CBS News — Foregrounds Justice Alito's recusal, his energy stock holdings, and detailed back-and-forth among justices during the Boulder climate case arguments.
Right · Fox News — Leads with conservative legal experts' interpretation that liberal justices sought a jurisdictional 'escape hatch' because they lack votes on the merits.
† Based on the headline and RSS summary only; the full article wasn't available.
Source: NYT (lean-left, nytimes.com), CBS News (lean-left, cbsnews.com), Fox News (lean-right, foxnews.com)
Weeks before Election Day, the FEC can't enforce the law from CBS News
The Federal Election Commission is down to just two members ahead of what's projected to be the most expensive election cycle on record.
Source: CBS News (lean-left, cbsnews.com)
Democrats now favored in key midterm races that could give them control of US Senate, new projections find – as it happened from The Guardian
This live blog is now closed.
Sign up for Breaking News US email
Two media giants, Paramount and Warner Bros Discovery, have completed their $110bn merger, creating a new entertainment powerhouse, Skydance.
The deal, fiercely contested by several state attorneys general over potential antitrust concerns, formally closed on Tuesday.
Source: The Guardian (lean-left, theguardian.com)
'Medicare for All' message attracts midterm voters buckling under medical bills from NPR
Midterm candidates are talking up Medicare for All, and in a poll 2 out of 3 Americans said they support it. The phrase speaks to a desire for radical change, though many don't really know what it is.
Source: NPR (center, npr.org)
Carpetbagger allegation, misplaced valor make Nevada's safe red House seat competitive from NPR
Northern Nevada has a reliably red congressional seat, represented by a Republican since it was drawn in the early 1980s. A complicated candidate and a potential blue wave this year could change that.
Source: NPR (center, npr.org)
Dems have edge in midterms, poll shows. And, Paramount-Warner Bros. merger to close from NPR
Polls are showing that Democrats have a clear advantage heading into the midterms, but that doesn't tell the whole story. And, today, a mega merger between Paramount and Warner Bros. is set to close.
Source: NPR (center, npr.org)
What voters in 6 key states could tell us about the 2026 midterms from NPR
Morning Edition is heading to Maine, Texas, North Carolina, Georgia, Ohio and Michigan to hear what voters say is shaping their choices ahead of the midterm elections.
Source: NPR (center, npr.org)
The Republican candidates walking a Trump tightrope from BBC News
Some running in November's elections are deciding whether to distance themselves from an unpopular president who still dominates his party.
Source: BBC News (center, bbc.co.uk)
What a post-midterm House Freedom Caucus could look like from The Hill
The House Freedom Caucus is heading into 2027 with a leadership shake-up and a new opportunity to shape how the influential conservative bloc operates, as several of its most prominent figures prepare to leave Congress. The caucus’s current policy chair, Rep. Chip Roy (R-Texas), and a number of other influential members, including GOP Reps. Ralph...
Source: The Hill (center, thehill.com)
‘Dumbest thing Democrats could do’ if they win the 2026 midterms is go after Trump, Patrick Bet-David tells ‘Pod Force One’ from New York Post
Patrick Bet-David warns Democrats that targeting President Donald Trump after the 2026 midterms could rally the MAGA base behind Eric or Lara Trump in 2028.
Source: New York Post (lean-right, nypost.com)
Trump attacks Democratic Senate candidate Annie Andrews with anti-trans slur from New York Post
Donald Trump used an anti-trans slur against South Carolina Senate candidate Annie Andrews while endorsing Republican incumbent Darline Graham.
Source: New York Post (lean-right, nypost.com)
First Maine Senate debate between Susan Collins and Troy Jackson
Left · NYT † — Frames the debate around Collins' midterm vulnerability and clashes with Jackson over Trump, ICE, and related issues.
Left · CBS News — Offers a detailed, balanced rundown of specific exchanges: Kavanaugh, the spending bill, tariffs, ICE, and each candidate's attacks.
Left · The Guardian — Foregrounds Collins distancing herself from Trump and Jackson's attacks, framing her as defensive amid a Democratic-leaning state. Left out: The Guardian did not mention Jackson's prior anti-abortion positions or Collins' attacks on his legislative record in Augusta.
Center · The Hill † — Highlights takeaways from a neck-and-neck race, emphasizing Collins spotlighting times she bucked her party.
Right · New York Post — Leads with Collins' apparent message that Trump should stay away, emphasizing her distancing from the president and attacks on Jackson's record. Left out: The Post did not mention Jackson's late replacement of Graham Platner after sexual misconduct allegations or the ICE shooting discussion.
† Based on the headline and RSS summary only; the full article wasn't available.
Source: NYT (lean-left, nytimes.com), CBS News (lean-left, cbsnews.com), The Guardian (lean-left, theguardian.com), New York Post (lean-right, nypost.com), The Hill (center, thehill.com), New York Post (lean-right, nypost.com)
Trump Does Not Plan to Reimburse Taxpayers for Promotional Ads
Left · NYT — Trump Does Not Plan to Reimburse Taxpayers for Promotional Ads: A day after the president said his super PAC would pay for the ads, which had drawn bipartisan backlash, officials said that he meant moving forward. A new taxpayer-funded ad aired Tuesday.
Left · CBS News — Trump says he'll start paying for ads after pushback over taxpayer funding: President Trump says he'll start paying for the pro-Trump ads that have blanketed the airwaves in recent weeks, using money from his super PAC instead of taxpayer dollars.
The side-by-side comparison wasn't available for this story; each outlet's own headline is shown instead.
Source: NYT (lean-left, nytimes.com), CBS News (lean-left, cbsnews.com)
Ex-CIA officer David Rush pleads guilty to defrauding government of millions
Left · NYT † — Foregrounds the plea hearing revealing fraud far larger than initially known, while stressing that key secrets in the case remain hidden.
Left · CBS News — Focuses on the guilty plea and forfeiture details, describing the fake classified programs used to divert government funds into Florida homes and gold. Left out: Rush's disclosure of a clandestine human source and classified information to a foreign government, reported by BBC.
Center · BBC News — Highlights the leaked-secret angle, including disclosure to a foreign government, alongside the fraud, restitution terms, and CIA Director Ratcliffe's response.
Right · Blindspot — No lean-right outlet in this digest's sources covered this story.
† Based on the headline and RSS summary only; the full article wasn't available.
Source: BBC News (center, bbc.co.uk), NYT (lean-left, nytimes.com), CBS News (lean-left, cbsnews.com)
Trump suggests Iran could 'take out' Los Angeles and San Diego at rally
Left · The Guardian — Framed the story around Newsom's 'treasonous' accusation, embedding it in a broader live politics blog with other US political news.
Left · NYT † — The RSS summary alone emphasized the bipartisan uproar and the White House claim that the remarks were taken out of context.
Left · CBS News — Foregrounded California officials' reactions and detailed the military and strategic importance of the two targeted cities. Left out: Newsom's 25th Amendment call was reported; CBS included it, but the Guardian's Collins and Rubio items were omitted.
Center · BBC News — Balanced presentation pairing Democratic condemnations with detailed White House defenses and background on both cities' military assets. Left out: Newsom's call for Trump's removal under the 25th Amendment was reported by CBS but not included.
Right · New York Post — Led with the White House's attack on the 'media freakout,' emphasizing Trump's framing and including personal jabs at Newsom. Left out: Newsom's directing the State Threat Assessment Center to monitor threats, reported by others, was omitted.
† Based on the headline and RSS summary only; the full article wasn't available.
Source: The Guardian (lean-left, theguardian.com), BBC News (center, bbc.co.uk), NYT (lean-left, nytimes.com), CBS News (lean-left, cbsnews.com), BBC News (center, bbc.co.uk), New York Post (lean-right, nypost.com)
An Influential Voice in the Climate Debate Has Died at 66
Left · NYT — An Influential Voice in the Climate Debate Has Died at 66: Joe Romm was a prolific writer who never hesitated to take on climate change deniers and the fossil-fuel industry — or even ostensible allies.
Left · CBS News — Dennis Hastert, disgraced former House speaker, dies at 84: Former Speaker of the House Dennis Hastert represented Illinois for two decades in Congress and was later jailed on charges related to his concealment of hush money payments made to cover up sexual abuse.
The side-by-side comparison wasn't available for this story; each outlet's own headline is shown instead.
Source: NYT (lean-left, nytimes.com), CBS News (lean-left, cbsnews.com)
Olympian wrongfully indicted in Reflecting Pool case seeks inspector general probes
Left · NYT † — Frames the story around Hearn's lawyers accusing prosecutors of ignoring evidence and providing blatantly false information about what they knew.
Left · CBS News — Detailed account of the IG request, the case's history, alleged prosecutorial misconduct by Pirro, and Trump's pressure on prosecutors.
Center · The Hill † — Leads with Trump's claim that pool repairs are nearly complete and his unfounded vandalism assertions, centered on his remarks.
Right · Blindspot — No lean-right outlet in this digest's sources covered this story.
† Based on the headline and RSS summary only; the full article wasn't available.
Source: NYT (lean-left, nytimes.com), CBS News (lean-left, cbsnews.com), The Hill (center, thehill.com)
Trump announces $6.6B Anduril shipyard in Baltimore for submarine components
Left · The Guardian — Foregrounded Palmer Luckey's Republican fundraising ties, the Trump administration's alignment with Anduril, and uncertainties around the government's claimed 40% stake. Left out: It left out China-related shipbuilding competition and the projected 3,100 direct and 11,000 indirect jobs that other outlets reported.
Left · NYT † — RSS summary only; framed the visit as a dispute over credit between Trump and Maryland's Democratic governor Wes Moore.
Right · Fox News — Framed the deal as a solution to the Navy's submarine production shortfall against China, detailing the commercial financing model and Anduril's technical plans. Left out: It left out Luckey's Republican fundraising history and Trump's unclear claim that taxpayers would receive a 40% stake.
Right · New York Post — Foregrounded Trump's leadership and 'peace through strength' rhetoric, including the sparring with Gov. Wes Moore over his invitation to the event. Left out: It left out Luckey's Republican fundraising history and the Guardian's note that Anduril has no prior experience building crewed submarines.
† Based on the headline and RSS summary only; the full article wasn't available.
Source: The Guardian (lean-left, theguardian.com), NYT (lean-left, nytimes.com), Fox News (lean-right, foxnews.com), New York Post (lean-right, nypost.com), New York Post (lean-right, nypost.com)
GOP Super PAC pours $5 million into South Carolina Senate race, sources say
Left · CBS News — GOP Super PAC pours $5 million into South Carolina Senate race, sources say: The influx of cash comes as President Trump escalated the attacks against Darline Graham's Democratic opponent, Annie Andrews.
Left · NYT — Republicans are spending to aid Darline Graham as the South Carolina Senate race tightens.
The side-by-side comparison wasn't available for this story; each outlet's own headline is shown instead.
Source: CBS News (lean-left, cbsnews.com), NYT (lean-left, nytimes.com)
Iran's Houthi allies deny losing ground as Mideast crisis engulfs Yemen from CBS News
Iran's Houthi allies deny that they've lost ground to Yemen's Saudi-backed military, as oil prices ease despite the Middle East crisis erupting in a battle over control of another key waterway.
Source: CBS News (lean-left, cbsnews.com)
Fort Hood shooter Nidal Hasan scheduled for December 3 execution by firing squad
Left · CBS News — Foregrounded victim notification and family impact, listing all thirteen victims' names and detailing Hasan's attack and his Fort Leavenworth death row status. Left out: It did not report Hasan's 2013 court-martial statements that he was protecting Muslims and the Taliban, which the BBC included.
Center · BBC News — Framed the firing squad as a historic first since World War Two, with broader context on military execution law and rarity of the method nationwide. Left out: It did not list the names of the thirteen victims or describe the notification process for their families as CBS did.
Right · Blindspot — No lean-right outlet in this digest's sources covered this story.
Source: CBS News (lean-left, cbsnews.com), CBS News (lean-left, cbsnews.com), BBC News (center, bbc.co.uk)
ABC seeks court order blocking FCC's early review of broadcast licenses
Left · CBS News — Foregrounded the courtroom arguments, quoting ABC's retaliation claims and the Justice Department's defense, plus the judge's skeptical questions. Left out: NPR reported Trump's prior $16 million settlement with ABC and broader context like the Kimmel suspension, which CBS left out.
Center · NPR — Framed the hearing as a wide-ranging First Amendment battle, adding background on Trump's grievances, the Kimmel suspension, and amici on both sides. Left out: NPR did not report the hearing's specific arguments about jurisdiction or Judge AliKhan's skeptical questioning of the Justice Department.
Source: CBS News (lean-left, cbsnews.com), NPR (center, npr.org)
Trump administration sends $90 Medicare and $500 ACA refund checks
Left · CBS News — Foregrounded expert skepticism that the checks fall short of rising healthcare costs and noted the payments' timing weeks before the midterms.
Center · The Hill † — Framed the payments as Trump's response to voter frustration over healthcare costs ahead of the midterm elections, with Republicans struggling to respond.
† Based on the headline and RSS summary only; the full article wasn't available.
Source: CBS News (lean-left, cbsnews.com), The Hill (center, thehill.com)
Skepticism mires Trump’s Iowa steel plant project: ‘Promises are not the same as jobs’ from The Guardian
The project has sparked doubts over whether the plant will be built and delivered as promised, given Trump’s record
Donald Trump has long sought to portray himself as a man of steel, from the White House photoshopping Trump to look like Superman, to enacting tariffs on steel and approving the sale of US Steel to Japan’s Nippon Steel with a ”golden” share given to the US government.
Most recently, the Trump administration announced plans to construct the largest steel plant in the US, a $15bn plant in Lee county, Iowa to be built by India-based Mesabi Metallics.
Source: The Guardian (lean-left, theguardian.com)
The surprising, cut-throat race for Los Angeles mayor
Left · The Guardian — The surprising, cut-throat race for Los Angeles mayor: Karen Bass is fighting for her political life as progressives battle to lead the US’s second-largest city Can the Iran war cost Trump in the US midterm elections? Your questions answered Karen Bass has never lost an ele…
Right · New York Post — Nithya Raman caught in ‘dangerous’ video targeting LAPD chief: ‘Fascist liars’: LA mayoral candidate Nithya Raman deleted a campaign video that branded Mayor Karen Bass and LAPD Chief Jim McDonnell 'fascist liars.'
The side-by-side comparison wasn't available for this story; each outlet's own headline is shown instead.
Source: The Guardian (lean-left, theguardian.com), New York Post (lean-right, nypost.com)
A history of presidents naming things after themselves (It's short. They rarely do) from NPR
While Trump pushes to get his name on the Kennedy Center and other public sites, history shows that sitting presidents have almost always avoided putting their brands on government assets.
Source: NPR (center, npr.org)
US death row inmate Christa Pike awake and speaking after failed execution, lawyers say from BBC News vs How did Christa Pike survive execution attempt - and what happens next? from BBC News
A medical expert advising her legal team says the murderer of Colleen Slemmer is communicating in a "basic way".
Source: BBC News (center, bbc.co.uk), BBC News (center, bbc.co.uk)
California pair charged with abusing at least 14 surrogate-born children from BBC News
Guojun Xuan and Silvia Zhang were arrested at their Arcadia mansion after 21 children were removed from their custody last year.
Source: BBC News (center, bbc.co.uk)
Lawsuit alleging drugged gang rape at Cornell fraternity prompts legal and political debate
Center · BBC News — Foregrounded legal questions about New York consent laws and the 'voluntary intoxication loophole', presenting expert and lawmaker perspectives on both sides. Left out: Did not mention political reactions blaming Trump's cultural influence or the criticism of the New York Times columnist.
Right · Washington Examiner † — RSS summary only
Right · Fox News — Foregrounded backlash against a New York Times columnist linking Trump to the alleged assault, emphasizing criticism and Trump's attack on Letitia James. Left out: Did not report details of New York consent law, the legislative loophole debate, or expert legal analysis.
† Based on the headline and RSS summary only; the full article wasn't available.
Source: Washington Examiner (lean-right, washingtonexaminer.com), BBC News (center, bbc.co.uk), BBC News (center, bbc.co.uk), Fox News (lean-right, foxnews.com)
Americans spend most on groceries in these 15 cities, study finds from The Hill
Inflation has backed off from the eye-watering double digit numbers after the pandemic, but unfortunately for Americans grocery store prices haven't gone down – an especially painful trend for residents of some U.S. cities, according to a new study.
Source: The Hill (center, thehill.com)
Why Trump chose these leaders to head his 'Super Intelligence Force' from The Hill
President Trump’s selection of four administration officials to head up his new “Super Intelligence Force” is offering a window into how the White House will approach AI in the face of pressure over its risks. The group of government leaders each reflect different priorities of the Trump administration after months of mixed messaging on the...
Source: The Hill (center, thehill.com)
GOP senators warn that Trump has fumbled immigration issue from The Hill
Republican senators say President Trump has fumbled the issue of immigration, something that was an advantage for GOP candidates in 2024 but is now a growing political liability for their party. Several GOP senators told The Hill that Trump's deportation policy has gone too far and warn that it's hurting Republican candidates with suburban and...
Source: The Hill (center, thehill.com)
Pentagon, under Hegseth, expands war against drug cartels in South America from The Hill
The Pentagon is expanding the U.S. military’s multifront war against drug cartels in South America. For more than a year, the U.S. military has been blowing up alleged drug boats on both sides of the continent. Forces have also begun interdicting refueling vessels off the coast of Ecuador before destroying them. And recently, U.S. special...
Source: The Hill (center, thehill.com)
AOC fundraiser remarks frame Abdul El-Sayed's Michigan Senate race
Left · Blindspot — No lean-left outlet in this digest's sources covered this story.
Center · The Hill † — RSS summary only; centers on the Michigan gubernatorial debate between Jocelyn Benson and John James, not the fundraiser story. Emphasis: the debate between Benson and James and its takeaways.
Right · Fox News — Foregrounds unearthed social media posts and DSA ties it says contradict El-Sayed's self-described capitalism. Left out: Did not report AOC's fundraiser remarks that El-Sayed's victory would signal left-wing wins anywhere.
Right · New York Post — Leads with AOC's private fundraiser comments, framing El-Sayed's race as part of a left-wing national strategy and her 2028 ambitions. Left out: Did not report the DSA posts about El-Sayed's capitalist labeling that Fox News highlighted.
Unrated · Freerepublic.com — Reposts a New York Times excerpt on AOC's remarks, followed by user comment threads criticizing AOC and El-Sayed.
† Based on the headline and RSS summary only; the full article wasn't available.
Source: Freerepublic.com (unrated, freerepublic.com), The Hill (center, thehill.com), Fox News (lean-right, foxnews.com), New York Post (lean-right, nypost.com)
Trump's daylight saving push gets new pressure campaign from House GOP leaders from Fox News vs Battle over daylight saving time erupts as push to stop changing clocks hits Senate wall from Fox News
Trump shared Tom Cotton's cell phone number after the Arkansas senator opposed the Sunshine Protection Act for permanent daylight saving time.
Source: Fox News (lean-right, foxnews.com), Fox News (lean-right, foxnews.com)
Ex-Obama spy chief prepped for anti-Trump trial by Biden-era special counsel's team, new records reveal from Fox News vs Biden-era FBI’s ‘Nixonian’ Trump family surveillance draws scrutiny from unlikely corner from Fox News
Declassified FBI memo reveals Jack Smith's team met with Trump-critic and former CIA Director John Brennan to discuss Trump's case, despite Smith telling GOP lawmakers he had never met Brennan before.
Source: Fox News (lean-right, foxnews.com), Fox News (lean-right, foxnews.com)
Congress killed the penny — now it's coming for the 13-cent nickel from Fox News
The Common Cents Act already targets the penny, but lawmakers now want the Treasury Department to find cheaper ways to produce the nickel coin.
Source: Fox News (lean-right, foxnews.com)
Federal judge blocks Trump administration fines of up to $1.8M for migrants from Fox News
A federal judge blocked the Trump administration's migrant fines of up to $1.8 million, ruling DHS unlawfully imposed civil penalties for ignoring final deportation orders.
Source: Fox News (lean-right, foxnews.com)
Wes Moore – potential 2028 White House hopeful – won’t commit to serving full term if re-elected Maryland governor from New York Post
“This governor will not commit to serving you, yet he’s asking for your vote," Gov. Moore's Republican opponent, Dan Cox, said.
Source: New York Post (lean-right, nypost.com)
Trump admin launches federal probe into alleged widespread ‘staff-on-student’ sexual misconduct at NYC public schools from New York Post
“Some of the reports indicate that there have been almost 2,000 instances in New York City, where school staff have engaged in sexual misconduct with students,” Assistant Secretary Kimberly Richey said.
Source: New York Post (lean-right, nypost.com)
Washington isn’t broken. It’s just run by insiders who don’t care about you from Washington Examiner
Public cynicism regarding Washington isn’t just a mood; it has become the default setting for American citizenship. When a tiny fraction of the electorate expresses genuine faith in our federal representation, we aren’t dealing with a simple policy disagreement. We are witnessing a wholesale failure of trust. Whether you sit on the Right, the Left, […]
Source: Washington Examiner (lean-right, washingtonexaminer.com)
Mamdani’s ‘hit list’ is pure ‘menis’ — and American politics is intoxicated by it from Washington Examiner
Back in the 1960s, when the Vietnam War was raging, a familiar scenario played out in small towns. During election years, you’d see a “Nixon’s the One!” sign planted right next door to a “McGovern for President” banner. Neighbors would be mowing the grass around their respective slogans, stop to wipe their brows, and wave […]
Source: Washington Examiner (lean-right, washingtonexaminer.com)
'People are overwhelmed': Community groups rally to inform voters ahead of Alberta referendum from CBC News
Politics professor Lori Williams and her colleagues are among the in-demand speakers as community-based organizations across Alberta plan, host and get creative to help residents understand the referendum's complex questions.
Source: CBC News (unrated, cbc.ca)
The obituary of Jeffrey Archer, written by Jeffrey Archer from Thedailymash.co.uk
JEFFREY Archer, 1940-2026, statesman, bestselling author, Olympic 100m sprint gold medallist, astronaut, and faithful husband has passed on. He sums up his life in his own words.
Source: Thedailymash.co.uk (unrated, thedailymash.co.uk)
Shyam Krishna Khadki registers candidacy for Nepali Congress president from Khabarhub.com
KATHMANDU: Shyam Krishna Khadki has registered his candidacy for president of the Nepali Congress at the party’s 15th General Convention. Khadki registered his candidacy to challenge party president candidate Gagan Thapa. Another Nepali Congress leader, Kisho…
Source: Khabarhub.com (unrated, english.khabarhub.com)
As Race Tightens, Republicans Resume Tepid Support For Max Miller in Ohio from NYT
The embattled congressman, accused of throwing scalding water at his ex-wife and breaking his daughter’s collarbone, rebuffed earlier Republican calls to drop out.
Source: NYT (lean-left, nytimes.com)
Vance says Iran must cut nuclear enrichment to end seven-month war from New York Post
VP JD Vance says Iran must reduce its nuclear enrichment capacity to end a seven-month war, as US officials remain unsure who is running Tehran.
Source: New York Post (lean-right, nypost.com)
