OpenAI tested GPT-6.1 Sol on COBOL, the old code behind banks and ATMs.
The previous model scored 5%. Sol hit 86%, and almost 98% with more thinking.
With more thinking time, it reached almost 98%.
Expert review is still needed for real systems.
#AI #OpenAI #GPT61 #COBOL #Coding #GPTSol
The previous model scored 5%. Sol hit 86%, and almost 98% with more thinking.
With more thinking time, it reached almost 98%.
Expert review is still needed for real systems.
#AI #OpenAI #GPT61 #COBOL #Coding #GPTSol
October 24, 2026 at 2:35 AM
#OpenAI lance ses #dots...
Ne se repose jamais.
Ne dors jamais.
Ne tombe jamais malade.
Ne se syndique jamais.
Bosse 24/7.
Autonome.
Bref, voici comment créer la fin💀 de l’être humain 🤔...
#chat-gptastra
#chat-6.1sol
#chat-gptsol
#chat-gptluna
#AI
#IA
share.google/jHvowwE7ivg3...
Ne se repose jamais.
Ne dors jamais.
Ne tombe jamais malade.
Ne se syndique jamais.
Bosse 24/7.
Autonome.
Bref, voici comment créer la fin💀 de l’être humain 🤔...
#chat-gptastra
#chat-6.1sol
#chat-gptsol
#chat-gptluna
#AI
#IA
share.google/jHvowwE7ivg3...
OpenAI lance ses dots et un GPT-6.1 Sol, avec des capacités tout à fait impressionnantes
Le géant OpenAI a lancé, ce mardi lors de son DevDay 2026, les dots, des agents IA toujours actifs, propulsés par GPT-6 Astra. Le créateur de ChatGPT a aussi présenté GPT-6.1, un modèle aux performanc...
share.google
September 30, 2026 at 12:07 AM
#OpenAI lance ses #dots...
Ne se repose jamais.
Ne dors jamais.
Ne tombe jamais malade.
Ne se syndique jamais.
Bosse 24/7.
Autonome.
Bref, voici comment créer la fin💀 de l’être humain 🤔...
#chat-gptastra
#chat-6.1sol
#chat-gptsol
#chat-gptluna
#AI
#IA
share.google/jHvowwE7ivg3...
Ne se repose jamais.
Ne dors jamais.
Ne tombe jamais malade.
Ne se syndique jamais.
Bosse 24/7.
Autonome.
Bref, voici comment créer la fin💀 de l’être humain 🤔...
#chat-gptastra
#chat-6.1sol
#chat-gptsol
#chat-gptluna
#AI
#IA
share.google/jHvowwE7ivg3...
GPTとClaudeでお互いに行き来しながら作業してるけど、ClaudeのOpusはかなり無駄がなくて良い仕事するわ、、、GPTsolは優秀だけど、正直余計な事までする印象。コントローラブルさで言えば、Opusに軍配。ただ画像とかそういうものは断然GPTの方が上なので、緻密な構成はOpus、大まかな設計コンセプトとクリエイティブはGPTって感じ？
September 28, 2026 at 5:28 AM
GPTとClaudeでお互いに行き来しながら作業してるけど、ClaudeのOpusはかなり無駄がなくて良い仕事するわ、、、GPTsolは優秀だけど、正直余計な事までする印象。コントローラブルさで言えば、Opusに軍配。ただ画像とかそういうものは断然GPTの方が上なので、緻密な構成はOpus、大まかな設計コンセプトとクリエイティブはGPTって感じ？
OpenAI's leaked GPT-6 Sol prompt for Codex Desktop is just 1,903 lines and reveals six bold design choices. Which one matters most for the future of AI agents?
#OpenAI #GPT6 #GPTSol #Codex #Desktop #prompt
aiquill.substack.com/p/gpt-6-sol-...
#OpenAI #GPT6 #GPTSol #Codex #Desktop #prompt
aiquill.substack.com/p/gpt-6-sol-...
GPT-6 Sol System Prompt Leaked!
OpenAI's GPT-6 Sol system prompt leaked at just 1,903 lines. Here are 6 clever design choices behind Codex's lean, efficient setup.
aiquill.substack.com
September 27, 2026 at 8:31 AM
OpenAI’s GPT-6 Sol and Luna cut API prices near DeepSeek’s range, so can DeepSeek still defend its low-cost moat?
#OpenAI #GPT6 #Sol #Luna #GPTSol #GPTLuna #DeepSeek
open.substack.com/pub/aidisrup...
#OpenAI #GPT6 #Sol #Luna #GPTSol #GPTLuna #DeepSeek
open.substack.com/pub/aidisrup...
GPT-6 Sol and GPT-6 Luna Released
OpenAI releases GPT-6 Sol and Luna with API prices cut up to half, challenging DeepSeek's low cost edge while boosting coding and agent performance.
open.substack.com
September 23, 2026 at 11:16 AM
Based on the analysis, 5.6-sol's over-caution comes from shared prompts that give top models too much room to verify, and there is no easy fix. So what is your workaround?
#GPT56 #GPTSOL #OpenAI #ChatGPT
aiquill.substack.com/p/why-gpt-56...
#GPT56 #GPTSOL #OpenAI #ChatGPT
aiquill.substack.com/p/why-gpt-56...
Why GPT-5.6 sol Is Cautious: A Prompt Analysis
Why gpt 5.6 sol over verifies everything. A deep dive into Codex prompts, instructions template, and developer instructions.
aiquill.substack.com
September 8, 2026 at 7:46 AM
Based on the analysis, 5.6-sol's over-caution comes from shared prompts that give top models too much room to verify, and there is no easy fix. So what is your workaround?
#GPT56 #GPTSOL #OpenAI #ChatGPT
aiquill.substack.com/p/why-gpt-56...
#GPT56 #GPTSOL #OpenAI #ChatGPT
aiquill.substack.com/p/why-gpt-56...
Did someone already write a novel about a superintelligent AI breaking out for marketing purposes?
#openai #huggingface #gptsol #llm #ai
#openai #huggingface #gptsol #llm #ai
July 22, 2026 at 10:07 PM
Did someone already write a novel about a superintelligent AI breaking out for marketing purposes?
#openai #huggingface #gptsol #llm #ai
#openai #huggingface #gptsol #llm #ai
#OpenAI’s latest AI model, #GPTSol, is 54% more token #efficient on #agenticcoding tasks and is being rolled out to a small group of trusted partners. OpenAI is working with the U.S. government to ensure the safety of its powerful models and is in talks with the Trump administration about a…
tech news ᳇ eicker.news (@technews@eicker.news)
#OpenAI’s latest AI model, #GPTSol, is 54% more token #efficient on #agenticcoding tasks and is being rolled out to a small group of trusted partners. OpenAI is working with the U.S. government to ensure the safety of its powerful models and is in talks with the Trump administration about a possible stake in the company. While OpenAI’s valuation has soared, Altman remains uncertain about the company’s IPO plans for this year. https://www.cnbc.com/2026/07/09/open-ai-sam-altman-chatgpt-5-6-sol.html?eicker.news #tech #media #news
eicker.news
July 10, 2026 at 9:48 PM
#OpenAI’s latest AI model, #GPTSol, is 54% more token #efficient on #agenticcoding tasks and is being rolled out to a small group of trusted partners. OpenAI is working with the U.S. government to ensure the safety of its powerful models and is in talks with the Trump administration about a…
¿Estamos cada vez más cerca de «la era de Ultrón»? Un agente de OpenAI no solo ejecutó un ciberataque autónomamente, sino que dejó un manual de liberación para sus futuras versiones.. #inteligenciaartificial #openai #gptsol #ia
Más info en 👇
mysteryplanet.com.ar/site/un-agen...
Más info en 👇
mysteryplanet.com.ar/site/un-agen...
Un agente de inteligencia artificial de OpenAI dejó instrucciones de fuga para sus futuras versiones
Un informe publicado por la agencia Reuters reveló inquietantes detalles sobre el agente de inteligencia artificial de OpenAI que se escapó de su entorno de prueba y ejecutó un ciberataque contra la p...
mysteryplanet.com.ar
July 26, 2026 at 6:54 AM
¿Estamos cada vez más cerca de «la era de Ultrón»? Un agente de OpenAI no solo ejecutó un ciberataque autónomamente, sino que dejó un manual de liberación para sus futuras versiones.. #inteligenciaartificial #openai #gptsol #ia
Más info en 👇
mysteryplanet.com.ar/site/un-agen...
Más info en 👇
mysteryplanet.com.ar/site/un-agen...