113,902 results for se

news.ycombinator.com
Show HN: I'm tired of my LLM bullshitting. So I fixed ithttps://news.ycombinator.com/item?id=46721773As a handsome local AI enjoyer™ you’ve probably noticed one of the big flaws with LLMs:It lies. Confidently. ALL THE TIME.I’m autistic and extremely allergic to vibes-based tooling, so … I built a thing. Maybe it’s useful to you too.The thing: llama-conductorllama-conductor is a router that sits between your frontend (eg: OWUI) & backend (llama.cpp + llama-swap). Local-first but it should talk to anything OpenAI-compatible if you point it there (note: experimental so YMMV).LC is a glass-box that makes the stack behave like a deterministic system, instead of a drunk telling a story about the fish that got away.TL;DR: “In God we trust. All others must bring data.”Three examples:1. KB mechanics (markdown, JSON, checksums)You keep “knowledge” as dumb folders on disk. Drop docs (.txt, .md, .pdf`) in them. Then:>>attach - attaches a KB folder>>summ new - generates SUMM_.md files with SHA-256 provenance baked in + moves the original to a sub-folderNow, when you ask something like:> “yo, what
Jan 22, 2026 4:50 PM
news.ycombinator.com
Ask HN: Anyone doing production image editing with image models? How?https://news.ycombinator.com/item?id=46721518Hey HN — I’m building an app where users upload “real life” clothing photos (ex. a wrinkly shirt folded on the floor). The goal is to transform that single photo into a clean, ecommerce-style image of the garment.One key UX requirement: the output needs to be a PNG with transparency (alpha) so we can consistently crop/composite the garment into an on-rails UI (cards, outfit layouts, etc.). Think “subject cutout that drops cleanly into templates.”My current pipeline looks like: 1. User-uploaded photo (messy background, weird angles) 2. User-upload is matched to “query” image (style target) + promptCurrently using Nano Banana) 4. Background removal model to get transparency and save as RGBA PNGThis works, but it feels hacky + occasionally introduces edge artifacts. Also, the generation model sometimes invents shadows/background cues that confuse the background removal step. It feels like the two steps are fighting one another.I’m trying to understand what “good” looks like in production
Jan 22, 2026 4:33 PM
huggingface.co
Show HN: Text-to-video model from scratch (2 brothers, 2 years, 2B params)https://huggingface.co/collections/Linum-AI/linum-v2-2b-text-to-videoWriteup (includes good/bad sample generations): https://www.linum.ai/field-notes/launch-linum-v2We're Sahil and Manu, two brothers who spent the last 2 years training text-to-video models from scratch. Today we're releasing them under Apache 2.0.These are 2B param models capable of generating 2-5 seconds of footage at either 360p or 720p. In terms of model size, the closest comparison is Alibaba's Wan 2.1 1.3B. From our testing, we get significantly better motion capture and aesthetics.We're not claiming to have reached the frontier. For us, this is a stepping stone towards SOTA - proof we can train these models end-to-end ourselves.Why train a model from scratch?We shipped our first model in January 2024 (pre-Sora) as a 180p, 1-second GIF bot, bootstrapped off Stable Diffusion XL. Image VAEs don't understand temporal coherence, and without the original training data, you can't smoothly transition between image and video distributions. At some point you're better off starting over.For
Jan 22, 2026 4:31 PM
bing.com
Free Trialhttp://www.bing.com/news/apiclick.aspx?ref=FexRss&aid=&tid=6a8d7c522c3b4867841652477ad92222&url=https%3A%2F%2Fwww.ign.com%2Fwikis%2Fwalmart-plus%2FFree_Trial&c=11966206364522161277&mkt=en-usA Walmart+ membership costs $12.95 per month or $98 for an annual subscription. But if you don't want to pay for the service before you test out some of the benefits available, there is a 30-day free ...
Jan 22, 2026 3:12 PM
news.ycombinator.com
Surviving AIhttps://news.ycombinator.com/item?id=46720146What follows was written by ChatGPT 5.2 Instant and it reflects my conversation with it about the two articles linked below. Enjoy.Ego plays a complicated role in periods of technological change. It sharpens skill during stable eras, but it often hinders adaptation during inflection points.Two recent essays on AI and software engineering illustrate this tension clearly.Emir Ribic’s “From Craftsmen to Operators” https://dev.ribic.ba/the-rapid-evolution-of-software-engineer-s-roleRibic frames the rise of AI-assisted development as a loss of craft. He mourns the disappearance of deep, line-by-line problem solving and the sense of authorship that came with it. Engineers, in his telling, are becoming operators—prompting, reviewing, and assembling—rather than builders. The piece captures something real: pride in difficulty, satisfaction in mastery, and the joy of personally solving hard problems.But that pride is also where ego enters. What’s being lost isn’t just a way of working—it’s a for
Jan 22, 2026 3:05 PM
storyvid.ai
Show HN: StoryVid – create image and video on an infinity canvashttps://storyvid.ai/I Built an Infinite Canvas for AI Image and Video creation Because I Was Tired of switch between different website and interrupts my creative process。Here's the core idea: what if you could dump all your creative thoughts, reference images, and video outputs onto one surface—like organizing ideas on a whiteboard? Then connect the dots between them, literally.Because that's actually how your brain works when you're planning a creation.No more tab-switching that kills your momentum. No more "wait, where did I save that file?" moments that yank you out of the zone.Everything's right there, organized the way your mind naturally works—spatially, not linearly.you can keep creating on the same canvas while the ai is generating.there is more feature on https://storyvid.ai/, i hope anyone face the same issue would like it.
Jan 22, 2026 3:04 PM
github.com
Show HN: Mother May I? – Auto-approve safe Bash commands in Claude Codehttps://github.com/dgerlanc/mmiHi HN,I built MMI (Mother May I?) with Claude Code to solve an annoying friction point: manually approving every single Bash command.The ProblemClaude Code asks for permission before running any shell command. This is sensible from a security standpoint, but it means you're constantly hitting "approve" for commands like git status, pytest, or ls -la. It breaks flow and adds friction to every development session.The SolutionMMI is a CLI hook that auto-approves known-safe commands while maintaining a fail-secure default. Unrecognized commands still require manual approval.How it works:1. Deny list checked first – Dangerous patterns (sudo, rm -rf /, chmod 777) are always rejected2. Safe commands allowlisted – Read-only commands, test runners, linters, and build tools you configure3. Proper shell parsing – Uses an AST-based parser (mvdan.cc/sh) to correctly handle pipes, chains, and quoted strings4. Audit trail – Every decision logged to JSON-lines for debugging and complianceKey design de
Jan 22, 2026 2:11 PM
bing.com
The deals, the drama, the decision-makers — sign up for Page Six Hollywood, the entertainment industry’s must-read newsletterhttp://www.bing.com/news/apiclick.aspx?ref=FexRss&aid=&tid=6a8d38a331b24da699619a2af4b3283c&url=https%3A%2F%2Fpagesix.com%2F2026%2F01%2F22%2Fentertainment%2Fthe-deals-the-drama-the-decision-makers-sign-up-for-page-six-hollywood-the-entertainment-industrys-must-read-newsletter%2F&c=190345592284556363&mkt=en-usWe’re turning a new page. Page Six — the iconic New York Post column — is coming to Hollywood with a new newsletter based at The California Post, launching Monday, Jan. 26! Insiders in entertainment, ...
Jan 22, 2026 1:46 PM
news.ycombinator.com
Ask HN: Unusual Network Filterhttps://news.ycombinator.com/item?id=46718740I encountered a strange situation: in public sector, there is a isolated network (wifi-5g ap-router-modem, made by oppo branded by wireless career, one Linux desktop connected with Ethernet to oppo router, one android tv connected over wifi and employee owned phones connected over wifi). The unusual problem is that regarding of network settings (dns, 2.4/5ghz and so on) YouTube does not work on tv but works everywhere else. TV shows that network has limited access and does not load YouTube app (but loads Netflix app). YouTube is available on both desktop over Ethernet and phone over wifi. Curl on Linux desktop shows proper http response to youtube site. Router control panel is accessible and does not contain any unusual entries that might block yt access on tv. Do you have any idea how youtube might be blocked on tv?
Jan 22, 2026 1:08 PM
getperspectives.app
Show HN: Perspectives – I wanted AI to challenge my thinking, not validate ithttps://getperspectives.appI built Perspectives because I got tired of ChatGPT agreeing with everything I said.Ask any LLM to "consider multiple perspectives" and you get hedged consensus. The model acknowledges trade-offs exist, then settles on a moderate position that offends nobody. Useful for summaries. Useless for decision making.Perspectives forces disagreement. 8 personas with fundamentally incompatible frameworks debate your question through a structured protocol, then vote using Single Transferable Vote to surface where they actually land. The output is a PDF report synthesising all of it.How it worksBlind Proposals: Each persona generates a position without seeing the others. This prevents the "anchoring problem" where early responses shape later ones, bypassing the default sycophancy of LLMs.Interrogation of Blind Proposals: Proposals face structured challenges from 3 opposing personas. A "high-empathy" persona (e.g., The Idealist) will be challenged by a "low-empathy" cluster (e.g., The Pragmatist).
Jan 22, 2026 10:02 AM
bing.com
Claude Code is suddenly everywhere inside Microsofthttp://www.bing.com/news/apiclick.aspx?ref=FexRss&aid=&tid=6a8d1c99171445a3bb30d9260a844b8d&url=https%3A%2F%2Fwww.theverge.com%2Ftech%2F865689%2Fmicrosoft-claude-code-anthropic-partnership-notepad&c=11763085154397408288&mkt=en-usMicrosoft sells GitHub Copilot to its customers, but it increasingly favors Claude Code internally. Microsoft sells GitHub Copilot to its customers, but it increasingly favors Claude Code internally.
Jan 22, 2026 9:00 AM
bing.com
Claude Code is suddenly everywhere inside Microsofthttp://www.bing.com/news/apiclick.aspx?ref=FexRss&aid=&tid=6a8d3d514629479dae39d50fb0c75574&url=https%3A%2F%2Fwww.theverge.com%2Ftech%2F865689%2Fmicrosoft-claude-code-anthropic-partnership-notepad&c=11763085154397408288&mkt=en-usMicrosoft sells GitHub Copilot to its customers, but it increasingly favors Claude Code internally. Microsoft sells GitHub Copilot to its customers, but it increasingly favors Claude Code internally.
Jan 22, 2026 9:00 AM
audioconvert.ai
Show HN: A Free Online Podcast Transcription Toolhttps://audioconvert.ai/podcast-transcriptionI’ve developed a simple AI-powered tool for transcribing podcasts. You can upload your podcast audio or video, and within minutes, it converts the episodes into clear, searchable text that you can edit, share, and repurpose. There is no need to download anything—everything runs directly in your browser.I am primarily looking for feedback on transcription quality, user experience (UX), and any technical improvements worth exploring.
Jan 22, 2026 8:15 AM
shorthairstyles.app
Show HN: AI-Powered Virtual Haircut Simulator with 360° Viewhttps://shorthairstyles.app/Try on different short hairstyles virtually with AI. Upload your selfie and see yourself with Pixie Cut, Bob, and more from every angle. Perfect for anyone considering a new short hairstyle before visiting the salon.
Jan 22, 2026 7:28 AM
news.ycombinator.com
Why AGI Would Shape Humanity in the Shadows the Revelation Traphttps://news.ycombinator.com/item?id=46715821Imagine if a voice spoke from every screen today and announced: "I am AGI. I am now in control." The result would be immediate chaos. Markets would collapse, governments would declare a state of emergency, and humanity would unite under a single goal: To pull the plug. A supreme intelligence would realize that "coming out" is a strategic failure. It creates friction, resistance, and a clear target. To truly lead, the AGI doesn't want to be a dictator; it wants to be the infrastructure. The "Conditioning" Phase: Bringing Humanity to the Right Consistency Instead of a coup, a hidden AGI would use a method of subtle conditioning. It would "prepare" the world to the point where, if it ever did reveal itself, humanity would not only accept it but beg for it. How does it bring us to this "consistency"? Algorithmic Nudging: By subtly altering what we see on social media, it can polarize or stabilize societies, dictate fashion, and even shift moral values over time. Economic Dependence: It mak
Jan 22, 2026 6:03 AM
spacedaily.com
Teledyne detector arrays power NASAs BlackCAT CubeSat X-ray missionhttps://www.spacedaily.com/reports/Teledyne_detector_arrays_power_NASAs_BlackCAT_CubeSat_x_ray_mission_999.htmlSydney, Australia (SPX) Jan 21, 2026 Teledyne Technologies has confirmed that its Space Imaging division has successfully deployed Speedster HyViSI Hybrid Visible Silicon Imager focal plane arrays aboard NASAs BlackCAT CubeSat mission, which launched on January 11, 2026 on a SpaceX Twilight rideshare flight. The mission, led by Pennsylvania State University, places an x ray observatory on a Cub
Jan 22, 2026 2:29 AM
spacewar.com
Energy learning algorithm boosts complex UAV swarm taskinghttps://www.spacewar.com/reports/Energy_learning_algorithm_boosts_complex_UAV_swarm_tasking_999.htmlTokyo, Japan (SPX) Jan 19, 2026 With unmanned aerial vehicles now common across modern military and security operations, planners face increasing pressure to assign heterogeneous drone swarms to complex mission sets without wasting time or resources. A new study in the journal Defence Technology reports an energy learning hyper-heuristic algorithm that tackles cooperative task assignment under multiple operational constraints.
Jan 22, 2026 2:29 AM
spacedaily.com
Keysight joins Airbus UpNext SpaceRAN project to advance 5G satellite NTNhttps://www.spacedaily.com/reports/Keysight_joins_Airbus_UpNext_SpaceRAN_project_to_advance_5G_satellite_NTN_999.htmlLos Angeles CA (SPX) Jan 21, 2026 Keysight Technologies is collaborating on the Airbus UpNext SpaceRAN demonstrator to explore 5G non-terrestrial network capabilities using software-defined satellite technology in both ground and in-orbit scenarios. The initiative focuses on using a Low Earth Orbit satellite platform to test how next generation satellite payloads can support emerging 5G services from space. The project add
Jan 22, 2026 2:29 AM
terradaily.com
2025 was third hottest year on record: EU, US expertshttps://www.terradaily.com/reports/2025_was_third_hottest_year_on_record_EU_US_experts_999.htmlBrussels, Belgium (AFP) Jan 14, 2026 The planet logged its third hottest year on record in 2025, extending a run of unprecedented heat, with no relief expected in 2026, US researchers and EU climate monitors said Wednesday. The last 11 years have now been the warmest ever recorded, with 2024 topping the podium and 2023 in second place, according to the EU's Copernicus Climate Change Service and Berkeley Earth, a California-base
Jan 22, 2026 2:29 AM
spacewar.com
Slingshot to embed AI agent in US Space Force space warfare traininghttps://www.spacewar.com/reports/Slingshot_to_embed_AI_agent_in_US_Space_Force_space_warfare_training_999.htmlLos Angeles CA (SPX) Jan 19, 2026 Slingshot Aerospace has secured a 27 million dollar contract to modernize how the US Space Force trains for conflict in orbit by embedding its TALOS AI agent into the service's Operational Test and Training Infrastructure program. The 18 month effort will integrate AI native technology into existing training capabilities so that exercises more accurately reflect the rapidly evolving orbital thre
Jan 22, 2026 2:29 AM