每家公司都有其独特的工作流程、工具、边缘情况和成功定义。(在新窗口中打开)Third Moment Research(在新窗口中打开)认为，最好的代理应该非常擅长那项特定工作，并且随着使用次数的增加，其能力应不断提升。
该公司正围绕这一理念构建一个自我改进的AI系统，利用OpenAI的最新模型来理解代理行为、识别反复出现的弱点，并围绕这些问题自动创建数千个合成场景。团队利用这些场景来改进代理，并随着新问题的出现重复这一过程。
Codex、GPT‑6 Astra、GPT‑6 Sol和GPT‑6.1 Sol支持该系统的研究和计算工作，使Third Moment能够同时运行10,000个代理，进行持续36至48小时的调查。该公司还使用Daybreak来测试代理工作环境的安全性。
“在单次会话中，GPT-6.1 Sol能够在数天的计算中保持连贯的计算研究思路，而不会偏离主线。”
——Adam Khakhar，Third Moment Research联合创始人
使用Codex减少运行时间和令牌浪费
Third Moment的早期应用之一是在量化金融领域，因为该领域的成功衡量标准清晰。这一领域为研究代理如何工作以及在哪里遇到困难提供了一个具体的环境。团队使用Codex来发现并解决研究环境中的问题。
在一个实例中，它使用Codex和GPT‑5.6（当时最新的GPT模型）来分析数千条记录的模拟轨迹（代理的动作和环境的响应）以及时间戳和代码库。
Codex发现了一个瓶颈：在一年期订单级美国股票数据上回测一个交易策略需要数小时。在等待期间，代理反复轮询完成状态，浪费了令牌，并让状态检查填满了它们的上下文。
在对回测进行性能分析并分析其输入和使用情况后，Codex识别出了一项缓存优化和一种内存打包技术。它试验了多种解决方案，并将修复部署到生产环境，将回测时间从约3小时缩短到约3至5分钟。这减少了令牌使用和上下文污染，为代理腾出更多令牌预算用于进一步实验。
使用Codex和GPT‑6让代理保持任务专注
Third Moment为每个代理提供特定工作所需的工具、信息和环境。对于量化研究，团队在Docker化环境中运行默认的Codex harness。代理会收到1TB的数据、文档，以及一个研究代码库，其中包含用于特征工程、GPU支持的模型训练、交易策略和模拟的模块。
GPT‑6 Astra、GPT‑6 Sol和GPT‑6.1 Sol提供推理能力，而Codex harness为代理提供工具，使其能够行动、创建计划、设定目标、生成子代理，并随着工作推进压缩上下文。
为了解决代理在完成任务前停止的问题，Third Moment将Codex视为由GPT‑6 Sol或GPT‑6.1 Sol驱动的外层harness中的子代理。当任务未完成时，外层harness会检查轨迹，并决定是重试、终止任务，还是升级给人工审查。
“Codex harness极其通用，使其能够跨许多不同领域使用。”
——Adam Khakhar，Third Moment Research联合创始人
在该研究环境中，代理致力于改进Third Moment工程师构建的基线交易系统。它们在历史数据上测试交易策略，评估某项改进是有意义的还是噪声，并利用每个结果来选择下一个实验。据Adam称，它们的工作改进了基线交易系统。
以更高的令牌效率维持更长时间的研究
Adam表示，在GPT‑5.5之前，模型只能执行用户指定的实验。借助GPT‑5.5和GPT‑5.6，与Codex harness配合工作的模型能够利用结果设计后续实验。如今，借助GPT‑6 Astra、GPT‑6 Sol和GPT‑6.1 Sol，代理可以通过与子代理并行运行的数百个实验来调查根本原因。
这一进步帮助代理探索了团队未曾预料到的研究方向。在一次会话中，GPT‑6.1 Sol识别出了非显示交易对某交易所的影响，这些交易在该交易所成交量中占比高得出乎意料。“这是一个不明显的结果，我们的研究团队最初并未考虑到，”Adam说。
Third Moment在优化、机器学习系统和量化研究方面评估模型。它在其最高推理设置下测试前沿发布版本，评估其浏览代码库、编写高效代码和持续进行调查的能力。当前测试还包括GPT‑6 Luna。
“鉴于其令牌效率，GPT-6.1 Sol目前是我们使用最多的模型。它以最少的令牌完成任务，同时提供与GPT-6 Astra相似的性能。”
——Jeffrey Yan，Third Moment Research联合创始人
一项基准测试要求代理降低实时记分牌服务的更新延迟，同时保持其输出不变。团队通过将专家手工解决方案的平均运行时间除以模型解决方案在相同硬件上的平均运行时间来衡量加速效果，并以正确性为前提。据Adam称，GPT‑6 Astra带来了最大的加速。
对于有针对性的工作，包括检查模拟系统中的特定安全弱点，团队使用更高效的模型，如GPT‑5.6 Terra。对于深度研究，Adam认为更强的推理能力可以证明更高成本是合理的：“在量化金融领域，发现某种有趣信号的价值极高。”
使用Daybreak自动化安全测试
在10,000个代理自主运行的情况下，Third Moment需要一种方法来大规模测试其研究环境的安全性。团队使用Daybreak网络安全平台在模拟启动前自动化该测试。
这些测试检查代理是否能够突破其研究环境并访问更广泛的网络。“通过使用Daybreak，我们能够让代理自身对模拟进行渗透测试，并测试其基础设施和安全性，”Adam说。
“如果没有Daybreak，我们将无法大规模自主测试这一点。”
——Adam Khakhar，Third Moment Research联合创始人
迈向数十万个协作代理
Third Moment的雄心延伸到公司每天所做的专业工作。对于零售商而言，这可能意味着在库存短缺或供应商延迟时做出更好的供应链决策。对于工程团队而言，这可能意味着在大型代码库中寻找性能瓶颈。对于文档密集型业务而言，这可能意味着识别一份文档中的承诺与另一份文档中记录的结果不匹配之处。
基于 GPT‑6 Astra、GPT‑6 Sol 和 GPT‑6.1 Sol 的推理能力，Third Moment 计划从个体研究扩展到数十万个协作子代理。其路线图要求数万亿个 token 支持数十个依赖实验的链条，目标是用这些研究在新挑战出现时自主改进代理。
对 Adam 来说，这将赋予人们更多能力来塑造代理的学习和工作方式：“人类可以处于一个更具杠杆作用的位置，他们可以思考如何进行研究并优化研究本身，而不是实际去做研究。”
Every company has its own workflows, tools, edge cases, and definitions of success.(opens in a new window)Third Moment Research(opens in a new window) believes the best agents should be really good at that specific work and should improve their capabilities the more they’re used.
The company is building a self-improving AI system around that idea, using OpenAI’s latest models to understand agent behavior, identify recurring weaknesses, and automatically create thousands of synthetic scenarios around those problems. The team uses those scenarios to improve agents, repeating the process as new issues emerge.
Codex, GPT‑6 Astra, GPT‑6 Sol, and GPT‑6.1 Sol support the system’s research and computational work, enabling Third Moment to run 10,000 agents at once on investigations lasting 36-48 hours. The company also uses Daybreak to test the security of the environments where agents work.
“In a single session, GPT-6.1 Sol is able to maintain a coherent thread of computational research across days of compute without losing the plot.”
—Adam Khakhar, Co-founder, Third Moment Research
Reducing runtime and token waste with Codex
One of Third Moment’s early applications is in the field of quantitative finance given the clear measure of success. This domain provides a concrete setting to study how agents work and where they struggle. The team uses Codex to find and address problems in the research environment.
In one instance, it used Codex with GPT‑5.6, the latest GPT model at the time, to analyze thousands of recorded simulation trajectories (the agents’ actions and the environment’s responses) alongside timestamps and the codebase.
Codex uncovered a bottleneck: backtesting a trading strategy across a year of order-level US equity data took several hours. While waiting, agents repeatedly polled for completion, wasting tokens and filling their context with status checks.
After profiling the backtest and analyzing its inputs and usage, Codex identified a caching optimization and a memory-packing technique. It experimented with solutions and deployed the fix to production, reducing the backtest from approximately 3 hours to approximately 3 to 5 minutes. This reduced token use and context pollution, freeing more of the agents’ token budget for further experiments.
Keeping agents on task with Codex and GPT‑6
Third Moment gives each agent the tools, information, and environment it needs for a specific job. For quantitative research, the team runs the default Codex harness inside a Dockerized environment. Agents receive a terabyte of data, documentation, and a research repository with modules for feature engineering, GPU-supported model training, trading strategies, and simulation.
GPT‑6 Astra, GPT‑6 Sol, and GPT‑6.1 Sol provide the reasoning, while the Codex harness gives agents tools to act, create plans, set objectives, spawn subagents, and compact context as work continues.
To address agents that stop before completing a task, Third Moment treats Codex as a subagent within an outer harness powered by GPT‑6 Sol or GPT‑6.1 Sol. The outer harness examines the trajectory when a task is left unfinished and determines whether to retry, terminate the job, or escalate to human review.
“The Codex harness is extremely versatile, and it enables use across many different domains.”
—Adam Khakhar, Co-founder, Third Moment Research
In that research environment, agents work to improve a baseline trading system built by Third Moment’s engineers. They test trading strategies on historical data, assess whether an improvement is meaningful or noise, and use each result to choose the next experiment. According to Adam, their work has improved the baseline trading system.
Sustaining longer research with greater token efficiency
Before GPT‑5.5, Adam says models could only execute experiments prescribed by a user. With GPT‑5.5 and GPT‑5.6, models working with the Codex harness could use results to design follow-up experiments. Now with GPT‑6 Astra, GPT‑6 Sol, and GPT‑6.1 Sol, agents can investigate root causes through hundreds of experiments running in parallel with subagents.
That progression has helped agents explore research directions the team hadn’t anticipated. In one session, GPT‑6.1 Sol identified the impact of non-displayed trades on an exchange where those trades accounted for an unexpectedly large share of volume. “This was an unobvious result that our research team did not initially consider,” says Adam.
Third Moment evaluates models across optimization, machine learning systems, and quantitative research. It tests frontier releases at their highest reasoning settings, assessing their ability to navigate repositories, write efficient code, and sustain an investigation. Current testing also includes GPT‑6 Luna.
“GPT-6.1 Sol is currently our most-used model given its token efficiency. It completes tasks with the fewest tokens while delivering performance similar to GPT-6 Astra.”
—Jeffrey Yan, Co-founder, Third Moment Research
One benchmark asks agents to reduce a live scoreboard service’s update latency while preserving its outputs. The team measures speedup by dividing the average runtime of an expert-crafted solution by that of the model’s solution on the same hardware, subject to correctness. GPT‑6 Astra delivered the biggest speedup, according to Adam.
For targeted work, including checks for specific security weaknesses in the simulation system, the team uses more efficient models like GPT‑5.6 Terra. For deep research, Adam believes stronger reasoning can justify higher cost: “In the space of quantitative finance, the value of finding some interesting signal is extremely high.”
Automating security testing with Daybreak
With 10,000 agents running autonomously, Third Moment needs a way to test the security of its research environments at scale. The team uses the Daybreak cybersecurity platform to automate that testing before simulations launch.
The tests check whether an agent could break out of its research environment and access the wider network. “By using Daybreak, we’re able to have agents themselves do penetration testing on the simulation and test the infrastructure and security of it,” says Adam.
“Without Daybreak, we would not be able to autonomously test this at scale.”
—Adam Khakhar, Co-founder, Third Moment Research
Moving toward hundreds of thousands of collaborating agents
Third Moment’s ambition extends to the specialized work companies do every day. For a retailer, that could mean making better supply chain decisions when inventory is short or a supplier is delayed. For an engineering team, it could mean finding performance bottlenecks in a large codebase. For a document-heavy business, it could mean identifying where commitments in one document don’t match outcomes recorded in another.
Building on the reasoning capabilities of GPT‑6 Astra, GPT‑6 Sol, and GPT‑6.1 Sol, Third Moment plans to expand from individual investigations to hundreds of thousands of collaborating subagents. Its roadmap calls for trillions of tokens supporting chains of dozens of dependent experiments, with the goal of using that research to autonomously improve agents as new challenges arise.
For Adam, this would give people more capacity to shape how agents learn and work: “Humans can be in a much more leveraged position, where they can think about how to do research and optimize the research itself instead of actually doing the research.”
本文内容采集自官方网站，排版和翻译可能与原页面存在差异。
阅读官方全文