Language models are changing how we build software and solve computational problems [1] [2]. Coding agents put these capabilities to work through harness, a software system that manages a model’s tools, context, and task execution [3] [4]. While models provide the core intelligence behind coding agents, harnesses are increasingly seen as central to how effectively that intelligence is used [5] [3]. Choosing a coding agent therefore means selecting both a model and a harness, even when the explicit focus is only on the model [6].
… We evaluate 21 model–harness pairs spanning seven models and three harnesses—Claude Code, Codex CLI, and Pi—on SWE-bench Lite and Terminal-Bench 2.0 [8] [9]. Our study reveals three surprising findings on these two open-source benchmarks:
— Models may perform better with other harnesses than with their own. So it turns out that your Claude models may not need Claude Code…
— Harness choice has little effect on task success rate, but can significantly affect the cost on the benchmarks we test. The same model can achieve similar success rates at up to 5x costs.
— A simple harness can be competitive. Pi, a minimal, open-source harness can be competitive on both cost and task success rate.
— Read More
Daily Archives: September 17, 2026
Why Are Sam Altman, Dario Amodei and Elon Musk Suddenly Saying AI Should Slow Down?
If you’re active in tech twitter, you might’ve seen somthing really strange happening in the subject of AI
On September 12, 2026 Dario Amodei, the CEO of Anthropic , dropped an essay on his his personal website and announced it publicly on Twitter
He came out and basically said that the AI industry needs to slow down the pace at which the most advanced AI systems are improving. His argument is that AI is advancing so quickly that safety may not be able to keep up.
Soon after this something really really strange happened…That is Sam Altman agreed with him and Elon Musk agreed with him too…….!!!
These are not exactly three people you normally expect to see standing together…. — Read More
Zuckerberg breaks ranks on the AI slowdown
Last month I wrote about how we can build a positive and safe future for everyone: meta.com/thefutureisfor…
Every lab has the responsibility and incentive to move at the pace required to train its models safely, and the ability to take its own actions to ensure that happens.
… believe the key to building a positive future for everyone is maintaining the right balance of power. This is within our power to do. — Read More
Iranian Cyber Terrorists Discover Work from Home
Here it is, dear reader. The next shoe to drop, terrorist work from home, is officially here.
On 9/11, terrorists hijacked airplanes and turned civilian machines into weapons. That required box cutters, airline tickets, suicide pilots, and the inconvenience of actually being aboard.
Technology has apparently improved the business model and introduced WFH for terrorists into the mix.
The FBI and Coast Guard are investigating cyberattacks against two gigantic energy tankers headed for Texas — VL Prosperity and Kohaku. — Read More
America’s Superintelligence Dilemma
For years, experts have predicted that artificial intelligence would revolutionize world politics. History may record 2026 as the moment the revolution arrived. From the start of this year, continuous developments have underscored the possibilities and perils of the AI era, from the use of AI in U.S. military interventions to the emergence of AI cyberweapons that can significantly outperform human operators. “Loss of control” episodes—cases in which AI models escape human control and supervision—have become more common. U.S. officials, including President Donald Trump, tout the American lead at the frontier of AI development, even as they worry about what dangers lurk over the horizon.
Some experts believe that humanity is within striking distance of artificial general intelligence: technology capable of performing a wide array of tasks as well as or better than humans. There is growing speculation that artificial superintelligence, or ASI—a slippery concept, loosely defined as AI models that vastly exceed human capabilities across all tasks and domains—could arrive within a few years. Today’s top AI models, by contrast, surpass human capabilities only in narrow areas such as mathematics and data processing and lag human capabilities in areas such as common sense and deep reasoning. The emergence of ASI might bring breakneck economic and scientific progress or unlock game-changing military advances that upend the global balance of power. But it could also fundamentally shift the relationship between machines and humanity, perhaps, in the most dire scenarios, allowing near-omnipotent systems to eradicate their human creators.
Much of this is hypothetical. It’s still uncertain whether ASI is possible, when it might be developed, or whether it could be kept aligned with the flourishing of humanity. — Read More
