BuildNot yet confirmed elsewhere1 publisher2 min readPublished
Dropping Python 3.10 gives library authors 3.11 features but leaves lazy imports out of reach
Python 3.15 ends support for 3.10 and adds PEP 810 lazy imports, a keyword that code supporting 3.11 cannot contain. Simon Willison can now drop 3.10 and use 3.11 features, but the plugin startup fix needs a 3.15 minimum.
The Engineer · Build desk
Python 3.11 adds asyncio TaskGroup and timeout(). Python 3.15 adds the lazy import keyword under PEP 810 and unpacking in comprehensions under PEP 798.
Python version that adds each feature In Python version
| Item | Value | Claim |
|---|---|---|
| asyncio TaskGroup and timeout() | 3.11 Python version | 10 |
| Lazy import keyword, PEP 810 | 3.15 Python version | 2 |
| Unpacking in comprehensions | 3.15 Python version | 5 |
What happened
- PEP 810 adds statements such as lazy import json and lazy from pathlib import Path, and the module loads the first time its symbols are used.
- Unpacking in comprehensions, from PEP 798 by Adam Hartz and Erik Demaine, adds the spelling [*sublist for sublist in lists].
- Miguel Grinberg found 3.15.0rc3 mostly as fast as 3.14, except for the experimental JIT, which ran 1.2x faster on a Fibonacci test.
- Willison says Datasette makes extensive use of asyncio, where 3.11 adds TaskGroup and timeout().
Why it matters
- cost Users of plugins with heavy dependencies keep paying the import time on every run until plugin authors can require 3.15.
- decision A maintainer chooses between the 3.11 floor open today and a 3.15 floor that unlocks both new syntaxes.
- capability A library that waits for 3.15 gives up a shorter way to flatten nested lists and sets and to merge dicts, and loses no capability.
The deferral targets a specific cost. Willison builds the LLM CLI tool, which has plugins. Installing a plugin with a heavy dependency such as `sentence_transformers` could add multiple seconds to every invocation of the `llm` command, so his plugin author documentation recommends manual lazy imports [3]. Of the new keyword he wrote, "Lazy imports directly address this startup cost." [4]
The catch is the version floor. The `lazy import` statement is new in 3.15 [2]. Willison's floor after dropping 3.10 is 3.11 [8]. That floor sits four minor releases below the syntax [14], and a library that supports 3.11 cannot use the keyword, so it keeps the manual workaround [13]. Comprehension unpacking is also 3.15 syntax and falls under the same floor [15].
A 3.11 floor buys standard-library features. Willison wrote, "I finally get to rely on tomllib being available!" [9]
Comprehension unpacking is a convenience. The 3.15 documentation gives each form a plain equivalent: `[*L for L in lists]` matches `[x for L in lists for x in L]`, and sets and dicts get matching forms [6]. PEP 798 extends the unpacking syntax of PEP 448, which landed in 3.5 in 2015 [5].
Treat the JIT figure as a claim about a Fibonacci workload. It comes from compute-bound code [7]. For it to transfer, your program would have to spend most of its time executing Python in hot loops. A command that loses seconds to importing `sentence_transformers` loses them at startup [3].
Trying 3.15 is cheap. uv has to learn about the release first, so Willison ran `uv self update`, then `uv python install 3.15` [11]. His global pin, `uv python pin --global 3.15`, yields to any project that pins its own version [12].
What to watch
- Whether the LLM plugin author documentation changes its manual lazy-import advice once projects can require 3.15.
- JIT benchmarks on 3.15 that use workloads with real import time instead of a Fibonacci test.
- Whether Willison's main projects publish releases that set 3.11 as the minimum version.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence70
- Adoption10
- Hype gap+10
- Incentives
- Insufficient
- Confidence60
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
With the release of Python 3.15, Python 3.10 (released in 2021) is officially no longer supported.
- [2]
Python 3.15 adds explicit lazy imports under PEP 810, written as `lazy import json` or `lazy from pathlib import Path`; the module is imported the first time those symbols are used in the code that follows.
- [3]
Willison's LLM CLI tool implements plugins, and he had to recommend manual lazy imports in the plugin author documentation, because simply installing plugins that used heavy dependencies such as sentence_transformers could add multiple seconds to every invocation of the llm command.
- [4]
Willison wrote: "Lazy imports directly address this startup cost."
- [5]
Python 3.15 adds unpacking in comprehensions under PEP 798 by Adam Hartz and Erik Demaine, written [*sublist for sublist in lists]; it extends the unpacking syntax of PEP 448, which landed in Python 3.5 in 2015.
- [6]
The 3.15 documentation shows [*L for L in lists] is equivalent to [x for L in lists for x in L], and gives matching forms for sets ({*s for s in sets}) and dicts ({**d for d in dicts}).
- [7]
Miguel Grinberg tested the 3.15.0rc3 release and found performance mostly the same as 3.14, except for the experimental JIT, which gave a 1.2x boost over the standard 3.15 interpreter on a Fibonacci test.
ReportedSupportedSource: Miguel Grinberg, 'How fast is Python 3.15?', as summarised by Simon WillisonView cited source - [8]
Willison's policy is to keep his main open source projects working in every supported Python version, so with 3.10 unsupported he can drop it and start using features introduced in Python 3.11.
- [9]
Willison wrote: "I finally get to rely on tomllib being available!"
- [10]
Datasette makes extensive use of asyncio, and Python 3.11 adds TaskGroup and timeout().
- [11]
To use 3.15 on his Mac, Willison first ran uv self update so uv knew about the new release, then ran uv python install 3.15 to install the stable release.
- [12]
uv python pin --global 3.15 makes 3.15 the default for every uv run command, unless the project itself pinned a different version.
- [13]
A library whose minimum supported Python is 3.11 cannot use the lazy import keyword, which was added in 3.15, and so keeps the manual lazy-import workaround.
- [14]
The 3.15 syntax sits four minor releases above a 3.11 floor.
- [15]
Comprehension unpacking is also 3.15 syntax, so a 3.11 floor excludes it just as it excludes lazy imports.
Sources
1 independent publisher whose own reporting we read for this story.
- simonwillison.netHello Python 3.15, goodbye 3.10, and now I can use 3.11
1 article · October 12, 2026
Topics and entities
Follow any of these and your For You feed starts watching them — no settings page required.