Sameh Nassar

How to Get Rich (without getting lucky):

@coryalthoff Elon Musk!

You just hit $1M ARR solo. Who is the first person you hire?

@sama In the end, it is just an AI tool that helps those who can use it the most and beyond just having a chat with it. Cobriums.com

so far at least, i'm pretty sure AI has been net job-creating. this was not what i expected--although i was much less pessimistic than others, i thought by this level of capability we'd have seen some impact. it is possible this direction keeps going!

@_iamhabeeb For sure.

Accept. Adjust. Adapt

@nikitabier Epic version for X! 😂💡🔥

Managing X when we reach AGI and have space datacenters

@sonialy0 Hosting on the Reputation! 😂

How did he deploy facebook without vercel ?

Meta muse Ain't playing🤬🤬

@RhysSullivan No kidding 😂

Anthropic is developing a new model

Ladys and gentlemen, This boy called by one of the largest tech company to work for it as Researcher now he is the chief Ai officer at @Meta and founder of the Scale AI. now he led the new AI Model ( Muse Spark 1.1 ) at @AIatMeta, @finkd 👏🔥✨

@leerob She seems to have found it by accident. 😂🫣

She’s found the ancient scrolls.

Get notifications for all of this account posts without Follow is a great idea besides to timeline customization 👏👍 #X

X The UI has been updated in a special way so you can choose/customize what you want to see only on your timeline instead of the randomness that existed before In addition, move the mailbox to people who aren't in your chat list so that you are aware of the conversations. This is useful because some users were avoiding it because of the scam/spam it had to avoid ignoring important messages. If you notice, the app opens 2x faster and the movement becomes smoother and more organized, and don't forget the glass façade. And the UI is different for Android users from iOS. Also, if you are a Grok user, you can create a video directly from photos ranging from 6-10 seconds for the video and above to come up. An important move from TeamX Led by @nikitabier @benjitaylor 👍👏🔥

@alexandr_wang Great model alex👏🔥

muse spark 1.1 is pretty good at frontend engineering

@Cristiano @selecaoportugal Goaaaat Ronaldo 🐐🌟🇵🇹⚽👑

Uma vitória de milhões!

@kanyewest Sorry! Who are you?!

I’m not working on a new album yet …

16/ Test GPT-5.6 Sol and ChatGPT Work on your longest current workflows. Measure token consumption, completion rate over extended sessions, and output fidelity against your templates. These are the concrete variables that determine value for builders and researchers. End of thread. 👏🌟💡 If you find this thread helpful plz put like, Retweet and Follow and join the Newsletter at Cobriums.com/newsletter.

15/ Key Takeaways for Evaluation GPT-5.6 Sol delivers 91.9% on Terminal-Bench 2.1, competitive cybersecurity results at ~1/3 token cost, and improved biology performance. ChatGPT Work adds sustained agentic execution across files and apps. The primary advantages are token efficiency in long tasks and reduced need for explicit step instructions.

#OpenAI released GPT-5.6 (Sol, Terra, Luna variants) and ChatGPT Work an agent powered by GPT-5.6 + Codex that acts across apps/files, sustains multi-hour projects, and turns goals into finished outputs (docs, decks, analyses, reports) matching your templates and style. This thread delivers the exact model specs, benchmark numbers, token efficiency data, rollout details, comparisons, and builder/research implications with no fluff. 1/ @OpenAI announced @ChatGPTapp Work: an agent powered by GPT-5.6 and Codex. It takes actions across apps and files, maintains project context for hours, and converts a high-level goal into finished outputs (documents, decks, analyses, sites, reports) while matching user templates, reference files, and style. 2/ GPT-5.6 Model Family Sol : Flagship tier. Highest performance in reasoning, coding, science, and cybersecurity. Terra: Balanced tier at lower cost. Luna: Fastest and lowest-cost tier. Sol powers the most advanced agentic features in ChatGPT Work. 3/ ChatGPT Work Core Mechanics Operates on web, mobile, and desktop. Uses context from selected apps and files. Executes multi-step workflows from goal descriptions without requiring step-by-step instructions. Generates style-matched outputs and sustains work over extended sessions. 4/ Terminal-Bench 2.1 Results (Agentic Coding) GPT-5.6 Sol Ultra scored 91.9%. This benchmark tests planning, iteration, and tool coordination in command-line workflows. It outperforms or matches top Claude models (Mythos 5, Fable 5) while using fewer output tokens in many cases. 5/ Cybersecurity Benchmarks On ExploitBench and ExploitGym (long-horizon vulnerability research and exploitation): GPT-5.6 Sol is competitive with leading models but uses approximately 1/3 the output tokens. This efficiency matters for sustained security analysis tasks. 6/ Biology & Science Workflows On GeneBench-style evaluations (genomics and quantitative biology): GPT-5.6 shows stronger results than GPT-5.5 while consuming fewer tokens. Supports longer-horizon scientific analysis pipelines with better token economy. 7/ Token Efficiency Data Across agentic coding, cybersecurity, and biology tasks, GPT-5.6 Sol frequently achieves frontier-level scores with substantially lower output token counts than comparable Claude models. This directly reduces cost and latency for long-running agent sessions. 8/ Advanced Modes “Max reasoning effort” setting for deeper step-by-step processing. “Ultra” mode for coordinating multiple agents on complex tasks. These enable handling of higher-complexity, multi-step objectives. 9/ Integration & Context Handling ChatGPT Work pulls live context from documents and apps, then produces outputs aligned to reference materials and preferred formats. This reduces manual context provision compared to standard chat interfaces. 10/ Rollout Details (as of July 9–10, 2026) Web and mobile: Available now for Pro, Enterprise, and Edu plans. Plus and Business plans: Rolling out over the following days. Desktop app: Merges Codex features with multi-tab support; available globally on Windows and Mac for all plans (including Free). Codex app users receive the update automatically. Follow the comments to learn more about GPT-5.6.