SShortSingh.
TopTechnologyAI & MLStartupsProgrammingScienceBusinessFinanceCrypto & Web3World
Back to feed
Programming

Terence Tao Writes on Lean Grimoire, a Resource for the Lean Proof Assistant

0
·2 views

A Hacker News post links to a blog entry by mathematician Terence Tao discussing Lean Grimoire. The article text provided contains only the link and discussion metadata, so the specific details of the project are not described in the source. The post received 5 points and no comments at the time it was captured. Lean is a proof assistant and programming language used for formal mathematics and software verification. Readers seeking the substance of the project would need to consult Tao's original post.

Read the full story at Hacker News

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Log in to join the discussion and vote.

Log in

Related stories

0
ProgrammingDEV Community ·

Five-Person AI Subscriptions Hinge on Break-Even Hours, Not Seat Price

A per-seat price does not capture the full cost of adopting an AI tool, nor show whether it pays off for a team. The author suggests the key test is whether a five-person team saves enough reviewed work time to cover the subscription and onboarding overhead. Using public US pricing checked on November 11, 2026, annual five-seat costs were roughly $1,200 for ChatGPT Business and $1,500 for Claude Team, with onboarding at an assumed $400 bringing first-year totals to $1,600 and $1,900. A simple break-even formula, monthly subscription cost divided by loaded hourly labor cost, shows ChatGPT Business needs about 3.13 team-hours saved per month. The author recommends a measured pilot on three recurring tasks, accounting for review time and failed attempts, alongside data, access, and integration considerations.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Claude Code can build an internal app and propose it to an osFoundry Room for approval

A guide explains how Claude Code can write a small internal tool, such as a standup timer, and propose it to a cloud Room in osFoundry. The proposal is not visible in the Room until the owner reviews the source files and clicks Approve and create. Once approved, the app appears as a tab in the Room's app bar that anyone in the Room can open. Claude Code connects to osFoundry through a login approved once per workspace, with read and write scopes, and it can verify results by reading app status. The Room must be a cloud Room assigned to a project the connection belongs to, and tool permissions can be restricted to read-only.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Eval gate passed at 97% because the agent's own answers were used as expected results

A support agent approved 37 refunds between $200 and $250 without human review over six weeks, violating a policy requiring approval above $200. A typo in a prompt edit had raised the limit in the agent's config, and the eval gate still passed 11 releases in a row. The gate missed the error because many expected answers had been recorded from the agent's own output, so the agent and the eval agreed with each other while both were wrong. The author concludes that expected answers must come from the policy or other source of truth rather than the agent, and that evals should not share the agent's config. After rewriting the cases from policy, seven previously encoded cases were found wrong and four recent releases failed, which had previously been considered good. The rebuilt set of 480 cases draws on production samples, policy boundaries, incidents, adversarial inputs and failure shapes.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Appliance firm loses small claims case after its AI gave conflicting warranty answers

A regional appliance repair company with six vans and an AI-first support inbox settled a small-claims case over a $740 compressor replacement last spring. A customer's 22-message, nine-day chat showed the AI giving four different warranty terms when she asked the same question repeatedly, including 12 months, 24 months, and a lifetime compressor warranty. The company argued its policy was 12 months, but the judge accepted the customer's reliance on the conflicting written statements. The article attributes the inconsistency to retrieval from different documents, including an outdated 2023 promo page and a manufacturer's parts sheet, rather than to the model simply making things up. Across one quarter, the company found 33% of threads where a repeated question produced at least two different answers.

0 comments Read more at DEV Community