aigest.dev
Today's digest · Thursday, September 10

The 13 things in AI/dev today.

LiveNext issue at 7:00 CET
Stories
13
93 filtered out
#1 / TODAY
Trail of Bits·2 min·20h agoFREE

A “proof” of Fermat’s Last Theorem that fits the margin

Trail of Bits reported a Lean bug affecting String.Pos.Raw.extract, Lean's low-level string-slicing function, in all stable versions up to 4.33.1, with a patch in v4.34.0-rc1. When asked to extract a one-byte slice at an astronomically large position, Lean's logical definition returns the empty string while compiled native code returns the entire original string, allowing a contradiction that yields a "checked" proof of Fermat's Last Theorem. The issue was found while using GPT-5.6 to experiment with a code-review skill.

Developers relying on Lean proofs should note that native_decide adds the compiler to the trusted boundary, so proofs can be accepted through tooling issues rather than kernel soundness.

leanformal-verificationsecurityproofs
blog.trailofbits.com
A “proof” of Fermat’s Last Theorem that fits the margin
ID design and primary keys
#2 / TOP STORY
LobstersFREE

ID design and primary keys

Alexey Makhotkin published part 1 of a series on primary key design at anchorsandlinks.com, described as a bonus chapter from the "Database Design Book." The post introduces external IDs, anchor IDs, and physical primary keys, arguing that key design should follow business requirements. It covers external IDs generated by outside systems, uniqueness constraints, and a small content management system example with an integer anchor table. Part 2 will address composite primary keys.

Infostealer Logs Expose Replayable AI Tokens That Can Bypass MFA
#3 / TOP STORY
The Hacker NewsFREE

Infostealer Logs Expose Replayable AI Tokens That Can Bypass MFA

The Hacker News reports that infostealer logs contain replayable AI tokens that can bypass MFA. The source states these tokens were found in infostealer logs and that they can be replayed to bypass multifactor authentication. The report frames the exposure as a consequence of credential-stealing malware capturing AI-related authentication material. The source does not name specific AI providers, token formats, or affected products.

aigest · daily

Get this every morning.

One email. The signal. Built for builders.

Free · Unsubscribe in one click · No trackers

// Worth knowing9 stories

Your work might not need the smartest model

The post's framing suggests developers should evaluate model choice against task requirements rather than defaulting to the most capable model.

microsoftmodel-selectionllmdevblogs
Microsoft Dev Blogs22h ago1mFREE

You probably do not need 264 AI agents

The excerpt contains only the 418 Challenge retro CSS, so developers get no source-grounded guidance on AI agent counts from this text.

dev-communitycssthemingai-agents
DEV Community3h ago1mFREE
More selected · 1
// Yesterday9 stories