What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Akka’s September 2026 experiment covered 65 open-source projects, but it did not produce 65 complete rewrites. The company reports that its first tranche took 99.3 hours and that 57 of the 65 ports improved either lines of code or performance. It then chose 10 projects for full implementation. The results describe one vendor’s workflow and selected projects—not proof that AI can reliably deliver or maintain arbitrary production software.
What did Akka test across 65 open-source projects?
In a report published September 3, 2026, Akka describes two stages of work. The first examined 65 projects and produced specifications and implementations covering up to 10% of each project’s surface area. Akka says the selection included projects that were poor fits for a port to Akka as well as projects that appeared more suitable.
As an Amazon Associate I earn from qualifying purchases.
The second stage selected 10 projects for full implementation, based on the company’s view of their potential impact and the availability of measurable baselines. So “65 projects” refers to a broad discovery and partial-implementation tranche; the full implementations were limited to a selected subset. Akka’s report describes the scope and method.
Free tools Windows power users keep installed
One-click scans. No signup required.
Akka reports that the initial tranche took 99.3 hours in total. It says 57 of 65 ports improved either lines of code or performance. That combined result does not mean all 57 improved both measures, or that every project was fully ported. InfoQ’s October 5, 2026 summary, drawing on Akka’s report, puts the initial tranche’s token use at 9.41 billion; that figure is reported by InfoQ rather than confirmed in the primary-source material summarized here. InfoQ’s summary provides its account of the token figure.
#1 Best Overall
How did the spec-driven workflow work?
Akka describes a recurring cycle of setup, discovery, porting, benchmarking, and improvement. The workflow treated specifications, tests, and audits as delivery work—not just preliminary prompts before code generation.
- Set up the project. The team prepared the source project and the environment needed to inspect and run it.
- Discover its behavior. Analysis of source code, domain models, schemas, and runtime behavior informed a specification of what the system did.
- Plan and implement the port. Akka Specify was used for planning, task breakdown, implementation, builds, tests, and review.
- Benchmark the result. A common runner compared test-suite execution, code size, and end-user latency.
- Use failures to refine the work. Test failures or gaps in the specification fed into subsequent iterations.
This process makes the specification and the criteria for passing consequential: an implementation can only be judged against behavior and limits that the team has made explicit. Akka says its report found that failures tended to occur when specifications left decisions implicit or auditors missed a class of error. That is the company’s interpretation of its own experiment, not an independently demonstrated causal finding. Akka’s report describes the workflow and its interpretation.
Did AI really port all 65 projects?
No—not in the sense of completing 65 equivalent, full rewrites. Most projects in the initial group received discovery work and implementations covering up to 10% of their surface area. Akka selected 10 for full implementation because it saw potential impact and measurable baselines.
The distinction matters when interpreting the headline metrics. The 57-of-65 figure is Akka’s reported count of ports with an improvement in at least one of two measures—lines of code or performance—within the initial tranche. It should not be read as 57 complete replacements, a 57-for-65 win on every benchmark, or evidence that all project behavior was preserved.
What did the experiment find about models, tokens, and code size?
InfoQ’s October 5, 2026 account of Akka’s findings reports differences in time and token use between Sonnet and Opus. These are secondary-source figures summarizing Akka’s work, not independently reproduced measurements.
| Reported comparison | What InfoQ reported | How to interpret it |
|---|---|---|
| Average time per port | Sonnet: 61 minutes; Opus: 120 minutes | InfoQ’s summary of the experiment; not a general model-speed benchmark. |
| Token consumption | Opus used about 40% fewer tokens than Sonnet | A reported comparison for this workflow, not a claim about token use across all tasks. |
| Effort settings | Higher effort increased consumption without consistently improving efficiency | InfoQ’s report of Akka’s findings; the summary does not establish a universal effect. |
The reported results do not identify a single model as an overall winner across quality, speed, cost, and project suitability. The available comparison is narrower: it describes time per port, token consumption, and effort settings as reported for this experiment. InfoQ’s account summarizes these figures.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What do the results establish—and what remains uncertain?
The experiment supports a limited conclusion: Akka reports that its delivery harness produced partial implementations across a deliberately varied set of projects and full implementations for 10 selected projects, with improvements on at least one of two tracked measures in 57 of the initial 65. It also illustrates a workflow in which specifications, tests, benchmarks, and auditor checks shaped the work.
It does not establish that autonomous AI delivery is broadly reliable across software projects. The initial tranche was mostly partial; the full implementations were selected rather than randomly assigned; and the evidence comes from the company whose SDK was used. The available accounts do not establish independent reproduction or a randomized control design, and they do not show that AI can maintain arbitrary production systems without human involvement.
Best Value
Akka’s report captures its own interpretation this way: “If there is a single thing to take from 65 ports, it is that the interesting variable in this system is not the model, not the effort, and not the runtime—it is the discipline of the specification and the auditors.” That is Team Akka’s conclusion about its experiment, rather than a neutral finding established across the software industry. The report contains the statement.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
