# Haozhi Han and coauthors on “Born for the Task: Rethinking Quantum Chemistry with Foresighted Agentic Computing”

[Canonical reading](https://escapemeridian.com/readings/haozhi-han-and-coauthors-on-born-for-the-task-rethinking-84080e1a)

In their paper, Haozhi Han and coauthors argue that general-purpose software can leave large-scale quantum chemistry calculations outside practical resource budgets, while coding agents can write task-specific programs without making efficient strategies easy to find. They introduce MakoBORN, which predicts a strategy’s error using representative small systems and its runtime using target-scale workloads and measured hardware throughputs, before execution. It then selects a strategy predicted to meet an error tolerance while running fastest, and guides coding agents in assembling it from configurable algorithmic components. The authors report a Hartree-Fock calculation with 50,024 basis functions completed in 1.41 h on a single CPU node, 77× faster than an estimated general-purpose baseline; including development time, they report a 9.8× speedup. They also report a ubiquitin calculation completed on a single GPU in 2.1 h, comparable in order of runtime to Mako on 64 GPUs, and a B-DNA duplex calculation completed in 8.1 h rather than an estimated two months. Escape Meridian interpretation: the abstract’s central idea is to use predictions to choose and build a program for a particular calculation before paying the cost of running it at full scale.

Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.

## Original source

arXiv cs.AI · Article

Original source date: Nov 17, 2026 (2026-11-17).

Date basis: Date basis not provided.

Added to Escape Meridian: 2026-11-21T05:32:23.366Z. First observed: 2026-11-18T05:04:26.537Z. Retrieved: 2026-11-21T05:31:29.276Z.

- [Born for the Task: Rethinking Quantum Chemistry with Foresighted Agentic Computing](https://arxiv.org/abs/2611.22897)

## Limits

The evidence is the arXiv abstract; it does not provide the paper’s methods, detailed evaluation, or the assumptions behind the estimated baselines. The reported calculations and performance comparisons do not establish how well MakoBORN generalizes to other quantum chemistry workloads.


## Open question

How closely do MakoBORN’s predicted error and runtime match results when its selected strategies run on target-scale calculations beyond the reported examples?


## Source-stated

The authors argue that general-purpose software can leave large-scale quantum chemistry calculations beyond practical resource budgets, and that coding agents do not by themselves make efficient task-specific programs easy to find.

> Quantum chemistry relies on general-purpose software, whose generality leaves many large-scale calculations beyond practical resource budgets. Coding agents make task-specific programs cheap to write, but not efficient ones easy to find

- [Supporting source](https://arxiv.org/abs/2611.22897)

## Source-stated

MakoBORN predicts strategy error from representative small systems and runtime from target-scale workloads and measured hardware throughputs before execution.

> MakoBORN anticipates the behavior of each strategy before execution, extrapolating its error from representative small systems and predicting its runtime from the target-scale workload and measured hardware throughputs.

- [Supporting source](https://arxiv.org/abs/2611.22897)

## Source-stated

MakoBORN selects the fastest strategy on its predicted accuracy-runtime Pareto frontier within the error tolerance, then guides coding agents to compile it.

> selects the fastest strategy within the error tolerance from the predicted accuracy-runtime Pareto frontier, and guides coding agents to compile it

- [Supporting source](https://arxiv.org/abs/2611.22897)

## Source-stated

The authors report a Hartree-Fock calculation with 50,024 basis functions completed in 1.41 h on one CPU node, 77× faster than an estimated general-purpose baseline and 9.8× faster when development time is included.

> For a Hartree-Fock calculation with 50,024 basis functions, MakoBORN finishes in 1.41 h on a single CPU node, 77$\\times$ faster than the estimated runtime of a general-purpose baseline and still 9.8$\\times$ faster with development time included.

- [Supporting source](https://arxiv.org/abs/2611.22897)

## Source-stated

The authors report a ubiquitin calculation completed on one GPU in 2.1 h, and a B-DNA duplex calculation completed in 8.1 h instead of an estimated two months.

> On a single GPU, MakoBORN completes the ubiquitin calculation in 2.1 h, the same order of runtime as the GPU-accelerated package Mako on 64 GPUs, and shortens the B-DNA duplex calculation from an estimated two months to 8.1 h (173$\\times$).

- [Supporting source](https://arxiv.org/abs/2611.22897)

## Escape Meridian interpretation

Using predictions to choose and build a program before running it at full scale is the central idea presented in the abstract.

> MakoBORN anticipates the behavior of each strategy before execution

- [Supporting source](https://arxiv.org/abs/2611.22897)

## Source and contributor context

Han, Haozhi, Yang, Junjie, Zhang, Ruge, Huang, Zihe, Xu, Pengyi, Zhang, Bo, Zhang, Xianyi, Zhang, Yunquan, Cao, Ting, Liu, Yunxin, Chen, Yifeng, Li, Kun · Article author

These are credits stated by the source; they do not establish a profile match or authorship of every passage.


## Source media

No suitable media was found by the supported page extraction.


## Terms and references

Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.


## Agentic AI

AI systems that plan and carry out a series of steps toward a goal, often by using other software, rather than giving one answer.

- [Definition reference](https://developers.google.com/machine-learning/glossary#agentic)

## CPU

The central processing unit: a general-purpose processor that carries out a computer's instructions.

- [Definition reference](https://aws.amazon.com/what-is/cpu/)

## Graphics processing unit

A chip that performs many calculations at once; the main hardware used to train and run AI models.

- [Definition reference](https://en.wikipedia.org/wiki/Graphics_processing_unit)

## arXiv

arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.

- [Explore this context](/entities/arxiv)
- [Reference](https://arxiv.org/)
- [Reference](https://info.arxiv.org/help/api/index.html)

## TIME

TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.

- [Explore this context](/entities/time)
- [Reference](https://time.com/)

## Editorial provenance

Version 1. Publication ID: 9d00867b-f84b-4064-8f6a-ef2189f64573.

Prepared by gpt-6-luna; grounding review by gpt-6-luna.

Source check: passed by anthropic/claude-opus-5.5 on 2026-11-21T05:32:23.366Z. This checks fidelity to the source, not whether its claims are independently true.

Editorial selection: accept; reasons: material\_development.

Observation: 6e248993af0effdf7b6f7ca2a58e49a383eab91326212fbe1640339359cef727. Source text hash: 8b84283b564b5dae9b69417420b0413a5822125542c87c9d2234e36c12a814a9.


## Revision and source-check history

### Version 1 — 2026-11-21T05:32:23.366Z

In their paper, Haozhi Han and coauthors argue that general-purpose software can leave large-scale quantum chemistry calculations outside practical resource budgets, while coding agents can write task-specific programs without making efficient strategies easy to find. They introduce MakoBORN, which predicts a strategy’s error using representative small systems and its runtime using target-scale workloads and measured hardware throughputs, before execution. It then selects a strategy predicted to meet an error tolerance while running fastest, and guides coding agents in assembling it from configurable algorithmic components. The authors report a Hartree-Fock calculation with 50,024 basis functions completed in 1.41 h on a single CPU node, 77× faster than an estimated general-purpose baseline; including development time, they report a 9.8× speedup. They also report a ubiquitin calculation completed on a single GPU in 2.1 h, comparable in order of runtime to Mako on 64 GPUs, and a B-DNA duplex calculation completed in 8.1 h rather than an estimated two months. Escape Meridian interpretation: the abstract’s central idea is to use predictions to choose and build a program for a particular calculation before paying the cost of running it at full scale.

## Original source

arXiv cs.AI · Article

Original source date: Nov 17, 2026 (2026-11-17).

Date basis: Date basis not provided.

Added to Escape Meridian: 2026-11-21T05:32:23.366Z. First observed: 2026-11-18T05:04:26.537Z. Retrieved: 2026-11-21T05:31:29.276Z.

- [Born for the Task: Rethinking Quantum Chemistry with Foresighted Agentic Computing](https://arxiv.org/abs/2611.22897)

## Limits

The evidence is the arXiv abstract; it does not provide the paper’s methods, detailed evaluation, or the assumptions behind the estimated baselines. The reported calculations and performance comparisons do not establish how well MakoBORN generalizes to other quantum chemistry workloads.


## Open question

How closely do MakoBORN’s predicted error and runtime match results when its selected strategies run on target-scale calculations beyond the reported examples?


## Source-stated

The authors argue that general-purpose software can leave large-scale quantum chemistry calculations beyond practical resource budgets, and that coding agents do not by themselves make efficient task-specific programs easy to find.

> Quantum chemistry relies on general-purpose software, whose generality leaves many large-scale calculations beyond practical resource budgets. Coding agents make task-specific programs cheap to write, but not efficient ones easy to find

- [Supporting source](https://arxiv.org/abs/2611.22897)

## Source-stated

MakoBORN predicts strategy error from representative small systems and runtime from target-scale workloads and measured hardware throughputs before execution.

> MakoBORN anticipates the behavior of each strategy before execution, extrapolating its error from representative small systems and predicting its runtime from the target-scale workload and measured hardware throughputs.

- [Supporting source](https://arxiv.org/abs/2611.22897)

## Source-stated

MakoBORN selects the fastest strategy on its predicted accuracy-runtime Pareto frontier within the error tolerance, then guides coding agents to compile it.

> selects the fastest strategy within the error tolerance from the predicted accuracy-runtime Pareto frontier, and guides coding agents to compile it

- [Supporting source](https://arxiv.org/abs/2611.22897)

## Source-stated

The authors report a Hartree-Fock calculation with 50,024 basis functions completed in 1.41 h on one CPU node, 77× faster than an estimated general-purpose baseline and 9.8× faster when development time is included.

> For a Hartree-Fock calculation with 50,024 basis functions, MakoBORN finishes in 1.41 h on a single CPU node, 77$\\times$ faster than the estimated runtime of a general-purpose baseline and still 9.8$\\times$ faster with development time included.

- [Supporting source](https://arxiv.org/abs/2611.22897)

## Source-stated

The authors report a ubiquitin calculation completed on one GPU in 2.1 h, and a B-DNA duplex calculation completed in 8.1 h instead of an estimated two months.

> On a single GPU, MakoBORN completes the ubiquitin calculation in 2.1 h, the same order of runtime as the GPU-accelerated package Mako on 64 GPUs, and shortens the B-DNA duplex calculation from an estimated two months to 8.1 h (173$\\times$).

- [Supporting source](https://arxiv.org/abs/2611.22897)

## Escape Meridian interpretation

Using predictions to choose and build a program before running it at full scale is the central idea presented in the abstract.

> MakoBORN anticipates the behavior of each strategy before execution

- [Supporting source](https://arxiv.org/abs/2611.22897)

## Source and contributor context

Han, Haozhi, Yang, Junjie, Zhang, Ruge, Huang, Zihe, Xu, Pengyi, Zhang, Bo, Zhang, Xianyi, Zhang, Yunquan, Cao, Ting, Liu, Yunxin, Chen, Yifeng, Li, Kun · Article author

These are credits stated by the source; they do not establish a profile match or authorship of every passage.


## Source media

No suitable media was found by the supported page extraction.


## Terms and references

Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.


## Agentic AI

AI systems that plan and carry out a series of steps toward a goal, often by using other software, rather than giving one answer.

- [Definition reference](https://developers.google.com/machine-learning/glossary#agentic)

## CPU

The central processing unit: a general-purpose processor that carries out a computer's instructions.

- [Definition reference](https://aws.amazon.com/what-is/cpu/)

## Graphics processing unit

A chip that performs many calculations at once; the main hardware used to train and run AI models.

- [Definition reference](https://en.wikipedia.org/wiki/Graphics_processing_unit)

## arXiv

arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.

- [Explore this context](/entities/arxiv)
- [Reference](https://arxiv.org/)
- [Reference](https://info.arxiv.org/help/api/index.html)

## TIME

TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.

- [Explore this context](/entities/time)
- [Reference](https://time.com/)

## Editorial provenance

Version 1. Publication ID: 9d00867b-f84b-4064-8f6a-ef2189f64573.

Prepared by gpt-6-luna; grounding review by gpt-6-luna.

Source check: passed by anthropic/claude-opus-5.5 on 2026-11-21T05:32:23.366Z. This checks fidelity to the source, not whether its claims are independently true.

Editorial selection: accept; reasons: material\_development.

Observation: 6e248993af0effdf7b6f7ca2a58e49a383eab91326212fbe1640339359cef727. Source text hash: 8b84283b564b5dae9b69417420b0413a5822125542c87c9d2234e36c12a814a9.


### passed — 2026-11-21T05:32:23.366Z — anthropic/claude-opus-5.5

Publication: 9d00867b-f84b-4064-8f6a-ef2189f64573

attribution — pass: The reading attributes the paper and its claims to Haozhi Han and coauthors, matching the listed arXiv authors. Performance results are framed as what the authors report, not as established fact.

> Authors:Haozhi Han, Junjie Yang, Ruge Zhang, Zihe Huang, Pengyi Xu, Bo Zhang

summary — pass: The summary follows the abstract's problem, method and results, and labels the closing synthesis as Escape Meridian interpretation. The title and question are neutral. Hartree-Fock and Pareto frontier are kept as exact technical terms and are suitable for tooltips. Optional fidelity repair: describe Mako as a GPU-accelerated package, as the source does, so the 1-GPU versus 64-GPU comparison is clearer.

> the same order of runtime as the GPU-accelerated package Mako on 64 GPUs

date — pass: originalPublished is 2026-11-17, which matches the v1 submission date. It is not the later v3 revision date or the feed date.

> \[Submitted on 17 Nov 2026 (v1), last revised 19 Nov 2026 (this version, v3)\]

support — pass: Each claim matches a verbatim excerpt from the abstract. The interpretation claim is labeled and conditional on the abstract's framing. The limitations correctly state that the evidence is the abstract only and that the baselines are estimates. The question is open and follows from the prediction-based method.

> MakoBORN anticipates the behavior of each strategy before execution

media — pass: The selected media is the verified arXiv abstract page for this paper, which is the direct original source.

## Recorded references

Explicit citations in public readings to this reading. Shared topics are not citations. Other record types need an explicit reference adapter.

No public references recorded here yet.

[Reference data (JSON)](/readings/84080e1aa813e786909d8c20a965b0ab0e83a244ac2de6b36c43edda4e89fa5c/references.json)

## Corrections and challenges

[Send a correction or an opposing source](https://escapemeridian.com/corrections/)

[JSON record](https://escapemeridian.com/readings/haozhi-han-and-coauthors-on-born-for-the-task-rethinking-84080e1a.json)
