Daily Brief: OpenAI launches GPT-5.3-Codex, most capable agentic coding model, as Codex App hits 1M downloads

OpenAI launches GPT-5.3-Codex, most capable agentic coding model, as Codex App hits 1M downloads

OpenAI dropped GPT-5.3-Codex on February 5, their most advanced agentic coding model yet, building on GPT-5.2 with 25% faster inference and top benchmarks like 56.8% on SWE-Bench Pro and 77.3% on Terminal-Bench 2.0.[1] It handles the full software lifecycle, from autonomous app and game building to cybersecurity hardening—earning the first "High capability" rating, complete with $10M in API credits for cyber defense.[1] The model even helped create itself, per OpenAI's cheeky quote.[1]

The new Codex App, launched February 2, racked up over 1M downloads in its first week with 60% week-over-week user growth, now available in CLI, Cursor, GitHub, and VS Code.[2][3] It's temporarily free for ChatGPT Free/Go users, with Sam Altman signaling plans to keep it accessible post-promo.[3] X is buzzing with viral demos of its agentic feats, fueling excitement for dev productivity boosts.

Perplexity upgrades Deep Research to Anthropic's Claude Opus 4.6, claims benchmark leadership

Anthropic unveiled Claude Opus 4.6 on February 5, packing upgrades in coding, agentic planning, and a 1M-token context window in beta, dominating benchmarks like Terminal-Bench 2.0 SOTA, 90.2% on BigLaw Bench, and leading GDPval over GPT-5.2.[4] Perplexity wasted no time, upgrading Deep Research for Max users ($167/mo) immediately, rolling out to Pro soon after, touting SOTA on Google's DSQA and internals with adaptive thinking and higher output tokens.[5][6]

Perplexity's move promises deeper research capabilities, as they claim: "Perplexity Deep Research now runs on Opus 4.6, improving our existing state-of-the-art results."[6] Max users on X are already raving about instant access and enhanced outputs, sparking talks of Perplexity pulling ahead in the benchmark wars.

X rolls out viral Grok-themed animation for like button

X flipped the like button into a spectacle around February 9, triggering a giant Grok logo animation on taps— a temporary gimmick that's got users spamming likes to show it off.[7][8][9] Echoing past hits like the SpaceX rocket effect, it's driving massive engagement in hours, with videos flooding feeds.

X reactions are pure hype: folks yelling "Hit the ❤️ button and see!" and thanking the team before it vanishes, turning mundane likes into a viral party.

Elon Musk warns of 'woke virus' in AI, prioritizes maximally truth-seeking AI for safety

Elon Musk resurfaced in a viral clip, slamming the "woke virus" in AI as a misalignment risk worse than HAL from 2001: A Space Odyssey, pushing for "maximally truth-seeking AI" as the ultimate safety play.[10][11] "My top concern for AI safety is that we need a maximally truth-seeking AI... It is very important to have truth," he said, reviving his TruthGPT crusade against biased models.

X is lit with shares and debates on truth vs. political correctness in alignment, underscoring ongoing tensions in AI ethics.