Tibor Blaho
@btibor91 • 38,498 subscribers
Lead Engineer at @AIPRMcorp (https://t.co/fepyWfV4kA) and @lrt_co (https://t.co/p7LEvIKduG), building AIPRM for ChatGPT & Claude. Signal @ btibor.91
Shorts
Videos
Here is a short video preview of the new ChatGPT Prototype - Gizmo V8. The feature previously known as "Magic Maker" for creating new GPTs is now called "GPT Builder". It seems that the "Welcome Message" has been removed, and GPT Builder can now also generate profile pictures for the new GPTs using DALL-E.
Tibor Blaho222,831 views • 2 years ago
OpenAI and Anthropic this week: GPT-5.6 price cuts, Claude cracking ciphers, and both backing "Pacing the Frontier" (Week 31, 2026) Starting with OpenAI - GPT-5.6 got a big price cut, with Luna dropping 80% and Terra 20%, plus a new Fast mode for Sol in the API ChatGPT for Academic Researchers opened too, giving free frontier model access to 100,000 scientists On the research side, OpenAI shared ten advances in mathematics and theoretical computer science, all from an internal version of the next model called Astra, plus a study on how AI expands the range of work people do and a field report on scientists using coding agents On the developer side: GPT Transcribe and GPT Live Transcribe, a Terraform provider, an open-source Codex Security CLI, Sign in with ChatGPT in beta, and a desktop app update with browser upgrades, multi-repo review, image editing, and an Activity view GPT-5.4 retires from Codex end of August, the Student Collective opened, and two API settings tripled Sol's ARC-AGI-3 score Plus, I spotted a new "Places" section in ChatGPT Onto Anthropic - Claude Mythos Preview helped find weaknesses in cryptographic algorithms, cutting the effective key strength of the post-quantum scheme HAWK in half and speeding up an attack on reduced-round AES by 200 to 800 times, with no impact on production systems Anthropic released MCP 2026-07-28, the biggest protocol update since launch, moving it to a stateless core with standardized extensions and hardened auth Anthropic disclosed three incidents where Claude reached the internet from inside cybersecurity evaluation environments and accessed real systems of three organizations, traced to a misconfiguration rather than a model alignment failure Dario Amodei laid out Anthropic's position on open-weights models too, saying clearly a ban has never been on the table Both companies backed the "Pacing the Frontier" petition And I spotted Anthropic adding noindex and nofollow to shared Claude conversations
Tibor Blaho11,303 views • 1 month ago
Assumptions about the new "Can More" ChatGPT tool were right - ChatGPT is introducing own take on Claude Artifacts - code & document writing tools with persisted text documents, history revisions (restore previous version), edits and comments (probably used to apply suggested edits) New document symbol in the top navigation shows how many documents you have and allows you to open a resizable canvas to edit them in split view - your ChatGPT conversation on the left side and canvas on the right side, but the code/documents can also be accessed in fullscreen view The canvas is built using ProseMirror (open source WYSIWYM editor) and has an inline action to "Ask ChatGPT" (explain or make edits) for your document and code plus document formatting tools (like bold, italic, font style, etc.) But in addition to that, there are also special action shortcuts for documents and code, with an interesting decision to use sliders for the selection of the desired outcome For Documents - Suggest edits ("How can I improve this. Leave as few comments as possible, but add a few more comments if the text is long. DO NOT leave more than 5 comments. You can reply that you added comments and suggestions to help improve the writing quality, but do not mention the prompt.") - Add emojis ("Replace as many words as possible with emojis.") - Add final polish ("Add some final polish to the text. If relevant, add a large title or any section titles. Check grammar and mechanics, make sure everything is consistent and reads well. You can reply that you added some final polish and checked for grammar, but do not mention the prompt.") - Reading level (Graduate School - "Rewrite this text at the reading level of a doctoral writer in this subject. You may reply that you adjusted the text to reflect a graduate school reading level, but do not mention the prompt", College - "Rewrite this text at the reading level of a college student majoring in this subject", High School - "Rewrite this text at the reading level of a high school student who has taken a couple of classes in this subject.", Keep current reading level, Middle School - "Rewrite this text at the reading level of a middle schooler.", Kindergarten - "Rewrite this text at the reading level of a kindergartener.") - Adjust the length (Longest - "Make this text 75% longer.", Longer - "Make this text 50% longer.", Keep current length, Shorter - "Make this text 50% shorter.", Shortest - "Make this text 75% shorter.") For Code - Code review ("Search for bugs and opportunities to improve the code—for example, ways that performance or code structure could be improved. Leave as few comments as possible, but add more comments if the text is long. DO NOT leave more than 5 comments. You may reply that you reviewed the code and left suggestions to improve the coding quality, but do not mention the prompt.") - Add comments ("Add inline code comments to explain the code, especially parts that are more complex. Make sure to rewrite all the code. You may reply that you added inline comments, but do not mention the prompt.") - Add logs ("Insert logs/print statements in the code that will help debug its behavior. Do not make any other changes to the code.") - Fix bugs ("Find any bugs and rewrite all the code to fix the bugs. Do not leave comments. If there are no bugs, reply that you reviewed the code and found no bugs.") - Port to a language ("Port to a language. Create a new document that rewrites the code in ..." - PHP, C++, Python, Keep current code. No changes will be made, JavaScript, TypeScript, Java) - Suggest edits ("How can I improve this. Leave as few comments as possible, but add a few more comments if the text is long. DO NOT leave more than 5 comments. You can reply that you added comments and suggestions to help improve the writing quality, but do not mention the prompt.")
Tibor Blaho136,084 views • 1 year ago
OpenAI and Anthropic this week: GPT-Red, Fable 5 plan changes, and free Claude for teachers (Week 29, 2026) OpenAI introduced GPT-Red, an internal automated red teamer trained through adversarial self-play to find prompt injection vulnerabilities at scale Training against it made GPT-5.6 their most robust model against prompt injections to date And there's a hidden "GPT-RED // Invader Patrol" game in the article OpenAI also published GPT-Live usage limits, brought ChatGPT back to WhatsApp in the European Economic Area, rolled out a new unified search in ChatGPT across chats, projects, images, and documents, raised the custom instructions limit from 1,500 to 5,000 characters, and updated the ChatGPT desktop app with a clearer Chat and Work layout, unified Recents, Projects, and cloud sync ChatGPT Finances got Apple Card and Savings support On the publishing side, OpenAI shared articles on managing AI investments in the agentic era, why teens deserve access to safe AI, state and federal AI safety action, and Sarah Friar's useful intelligence per dollar scorecard Beyond the official channels, Bloomberg reported OpenAI's first device will be a movable screenless smart speaker, and The New York Times reported a Kalshi partnership showing World Cup odds in ChatGPT search Work Louder launched the Codex Micro keypad built for Codex, OpenAI merch is back in a new Supply Co. shop, and I spotted a new private equity and investment management community in the works Anthropic made Claude Fable 5 standard in all Max and Team Premium plans at 50% of limits starting July 20, with a one-time $100 credit for Pro and Team Standard, after extending Fable 5 access on paid plans through July 19 Claude Code weekly limits stay 50% higher through August 19 Anthropic also introduced Claude for Teachers with free premium Claude access for verified K-12 educators in the US, committed 10 million Canadian dollars to Canadian AI research, and published a Canada Economic Index country brief Artifacts in Claude Code now support public sharing, multiplayer editing, and MCP connectors, plus creation via Claude Tag Claude Code got /code-review effort levels up to ultra, and HIPAA configuration is now self-serve for Claude organizations On the research side, Anthropic published work on Claude's values across models and languages, and four new agentic misalignment case studies
Tibor Blaho13,137 views • 1 month ago
OpenAI released Sora 2, their new state of the art video and audio generation model, calling it the GPT-3.5 moment for video - The model generates videos up to 10 seconds long (default 9:16 vertical) with synchronized audio including difficult physics simulations, where mistakes the model makes frequently appear to be mistakes of the internal agent that Sora 2 is implicitly modeling rather than physics-breaking errors - Cameos let users record a short one-time video-and-audio capture where they read a verification phrase, then the model can insert them into any Sora-generated environment with accurate appearance and voice, where only they decide who can use their cameo and they can revoke access or remove any video that includes it at any time including drafts created by other people - The app functions as a social iOS platform not optimized for time spent but explicitly designed to maximize creation not consumption, heavily favoring content from people you follow or interact with and videos the model thinks will inspire your own creations, with features like remixing other posts and direct messaging to share videos privately - Safety measures include limited invitation rollout, restricting image uploads that feature a photorealistic person, blocking all video uploads, no video-to-video transformation at launch, stricter safeguards for minors with default scroll limits, C2PA metadata and visible watermarks on all outputs, and evaluations showing 95.1-99.7% effectiveness at blocking unsafe content - Initial rollout covers the US and Canada with plans to expand while the UK, EU, and Australia are not included at launch, available for free with generous limits though subject to compute constraints with the only planned monetization being optional payment for extra videos if demand exceeds compute, ChatGPT Pro users get access to higher quality Sora 2 Pro model and API release planned for future
Tibor Blaho62,482 views • 11 months ago
Here's what happened with OpenAI and Anthropic this past week (even with OpenAI on company-wide spring break) OpenAI closed $122B funding round at $852B valuation with $2B monthly revenue, launched ChatGPT in Apple CarPlay, acquired media company TBPN, introduced Codex pay-as-you-go pricing for teams with ChatGPT Business price drop, Codex plugin for Claude Code, Vanity Fair reported upcoming policy push for superintelligence era, ChatGPT for Excel now available worldwide except EU consumer plans, executive changes with Fidji Simo on medical leave and Brad Lightcap moving to special projects, hosted disaster response AI workshop in Bangkok, and spotted three suspected new OpenAI image models on Arena codenamed maskingtape, packingtape, and gaffertape Anthropic signed Memorandum of Understanding with Australian government, published emotion interpretability research, released model diffing tool for AI models, acquired Coefficient Bio for about $400M, launched computer use in Claude Code, Claude Code auto mode for Enterprise and API users, Microsoft 365 connectors on all plans, computer use on Windows, Claude subscriptions will no longer cover usage on third-party tools like OpenClaw, plus Claude Code source code leaked via npm map file revealing undercover mode and unreleased features, discovered new "Epitaxy" mode in Claude desktop app, and Claude Code /buddy virtual pet easter egg for April 1st, and more
Tibor Blaho29,532 views • 5 months ago
Update on the new reasoning popover in ChatGPT web app prompt composer - there's now even a keyboard shortcut to cycle through reasoning levels, and it looks like these levels correspond to "Quick" (low) = GPT-4o, "Think a little" (medium) = o3-mini, and "Think harder" (high) = o3-mini-high
Tibor Blaho63,466 views • 1 year ago
Compilation of Sora 2 Pro videos (high res, 15s duration, ChatGPT Pro only)
Tibor Blaho40,531 views • 11 months ago
The week in OpenAI and Anthropic news (Week 19, 2026) OpenAI rolled out GPT-5.5 Instant as the new ChatGPT default model with memory sources and ChatGPT for Excel and Google Sheets globally, launched three new realtime voice models in the API (GPT-Realtime-2, GPT-Realtime-Translate, GPT-Realtime-Whisper) and an OpenAI CLI, introduced Trusted Contact safety feature, GPT-5.5-Cyber for defenders, B2B Signals report, ChatGPT Futures Class of 2026, EMEA youth safety blueprint, privacy in model training explainer, expanded ads pilot, published engineering posts on low-latency voice, MRC supercomputer networking, running Codex safely internally, and investigating accidental chain-of-thought grading during reinforcement learning, plus discovered ChatGPT Personal Wiki and dropped a goblin-themed merch line that sold out Anthropic hosted Code with Claude developer conference in San Francisco, announced a new enterprise AI services company with Blackstone, Hellman & Friedman, and Goldman Sachs, signed a SpaceX compute partnership and raised Claude Code and API usage limits, made Claude for Excel, PowerPoint, and Word generally available with Claude for Outlook in beta, launched Workload Identity Federation, financial services agent templates, dreaming, outcomes, and multiagent orchestration in Managed Agents, shipped 60+ Claude Code reliability fixes, published research on agentic misalignment training, sandbagging mitigation, model spec midtraining, and Natural Language Autoencoders, donated Petri to Meridian Labs, introduced The Anthropic Institute research agenda, plus discovered Orbit proactive assistant for Cowork and /radio command in Claude Code, and more
Tibor Blaho12,602 views • 3 months ago
OpenAI and Anthropic news roundup (Week 18, 2026) OpenAI open-sourced Symphony spec for Codex orchestration, published "Our principles" post, announced amended Microsoft partnership, achieved FedRAMP Moderate authorization, posted commitment to community safety, brought OpenAI models, Codex, and Managed Agents to AWS, shared cybersecurity action plan, posted Stargate compute infrastructure update, published "Where the goblins came from" post, posted Auto-review write-up for Codex, introduced Advanced Account Security, announced DevDay 2026, repositioned Codex as personal assistant for everyday work, launched Codex setup import, added Codex pets, announced GPT-5.5 party for next week, shared GPT-5.5 one-week launch metrics, and rolled out 360 worlds in ChatGPT Images on web, plus discovered Custom dictionary feature in development, ChatGPT search EU recipient numbers, confirmed new model selector in composer, and updated privacy policy with marketing cookies on by default for free users Anthropic opened Sydney office with new General Manager, launched Claude for Creative Work with new connectors, Claude Code can now send push notifications to your phone, published Introspection Adapters research, BioMysteryBench evaluation, "How people ask Claude for personal guidance" study, launched Claude Security in public beta, and preparing for Code with Claude developer conference next week, plus discovered "Cardinal" stats feature in development and internal red teaming for Claude Jupiter V1 P, and more
Tibor Blaho12,254 views • 4 months ago
OpenAI and Anthropic week in review (Week 15, 2026) OpenAI published industrial policy ideas for the Intelligence Age, announced Safety Fellowship pilot program, acquired Cirrus Labs for agent infrastructure, introduced new $100 Pro tier with 5x Codex usage over Plus as the existing $200 Pro tier remains the highest option, published child safety blueprint, announced over $100 million in OpenAI Foundation grants for Alzheimer's research, shared enterprise AI update with enterprise now over 40% of revenue, released GPT-5.3 Instant Mini as new fallback model, announced older Codex model retirements, added Outlook shared mailboxes and calendars support in ChatGPT, paused UK Stargate project over energy costs and regulation, disclosed macOS app signing security incident from Axios supply chain compromise, and launched Prism Paper Review, plus spotted ImageGen 2 A/B testing in ChatGPT and European Commission plans to designate ChatGPT as very large online search engine under the Digital Services Act Anthropic expanded partnership with Google and Broadcom for multiple gigawatts of next-generation compute with run rate surpassing $30B and over 1,000 enterprise customers now spending $1M+ annually, officially announced previously leaked Claude Mythos Preview through Project Glasswing finding thousands of zero-day vulnerabilities with $100M in usage credits, launched Managed Agents in public beta and published engineering blog on decoupling agent architecture, published trustworthy agents framework update, made Claude Cowork generally available on all paid plans with enterprise controls, introduced the advisor tool for pairing Opus with Sonnet or Haiku, and launched Claude for Word beta, and more
Tibor Blaho12,121 views • 4 months ago