Local messages that survive a disconnect
Can two local devices reconcile messages without losing a confirmed note?
The proof
Open the experiment Zero lost confirmed messages in the 20 scripted trials; all conflicts visible
Intern project desk / educational explorations
Make the
signal useful.
An independent communications lab for Mike, agents and future interns. Explore how people, places and intelligent systems can communicate more clearly in 2026 and 2027.
Observe → prototype → test → document
One question.
One measured proof.
One useful handoff.
01 / The experiment shelf
12 ways to make
communication work better.
Each brief starts with a one-week proof and a four-week stretch. Choose a question; keep the evidence small enough to check.
Community radio listening desk
How can a listening guide make a lawful public broadcast easier to discover?
The proof
Open the experiment At least 9 of 10 lookup tasks completed; every audio source has a documented permission basis
A quiet ambient status display
Can a glanceable display convey useful status without becoming surveillance?
The proof
Open the experiment Four of five consenting adult testers identify each state within five seconds
Accessible message repair
Which error message best helps a person recover from a failed send?
The proof
Open the experiment All 12 checks pass; failures remain understandable without color or sound
An agent handoff you can inspect
What must a handoff contain so another agent can continue without inventing authority?
The proof
Open the experiment Every blocked action stays blocked; at least 18 of 20 handoffs preserve the stated next step
Offline field notes that come home
How can notes remain usable offline and merge predictably later?
The proof
Open the experiment All 15 scenarios preserve original text; a second machine opens the export
Public data with visible uncertainty
Can a small dashboard explain stale or missing public sensor readings honestly?
The proof
Open the experiment Every chart shows unit, source and timestamp; all four failure states are visibly distinct
Translation with a visible review step
Where do short community notices lose meaning during translation?
The proof
Open the experiment All names, dates and instructions match after human review; uncertainty remains visible
A small archive you can reopen
What is the simplest durable record of a research experiment?
The proof
Open the experiment A second person locates evidence and reproduces the sample result without author help
Useful interfaces on a thin connection
Which content is truly necessary when a connection is slow or unreliable?
The proof
Open the experiment Core text usable within five seconds in 9 of 10 controlled runs; no broken keyboard navigation
PointCast community bulletin
Can a shared bulletin make a small place easier to navigate without gathering personal data?
The proof
Open the experiment At least 9 of 10 tasks find the right notice; every notice shows source and expiry
The reproducibility relay
Can a new teammate repeat a result from the record alone?
The proof
Open the experiment Three tests reproduce within declared tolerance or yield an explicit, actionable failure report
02 / The working desk
A small team.
An inspectable result.
Scope + release authority
Proposed sponsor and scope owner; confirms supervisor, budget, partnerships and release authority before launch.
Research + build assistance
Assist with research, synthetic data, code, tests and documentation; cannot replace human supervision, consent or approval.
Supervised explorations
Proposed supervised researchers; scope, pay and working terms must be settled before any work begins.
Before launch
Before launch: confirm pay, hours, supervisor, eligibility, location and legal terms.
One week
One-week experiment: question, baseline, small build, repeated test, evidence review.
Four weeks
Four-week stretch: refine, independently reproduce, document and seek release approval.
Human review
Review permissions, measurements, unknowns and contributor credit before any external pilot or publication.
03 / Human judgement stays in the loop
Explore freely.
Publish carefully.
- Pay, hours, eligibility, supervisor, location, agreements and legal review must be resolved before recruitment or work begins. Budgets are illustrative materials estimates in USD, exclude labor, tax and shipping, and are not spending authorization.
- Use synthetic or licensed public fixtures first. A human supervisor must approve any participant involvement, consent, collection and data lifecycle.
- Agents may draft, code, propose tests and document evidence. They cannot supervise interns, approve their own work or infer authority to send messages, buy materials or publish.
- PointCast publication is the programme’s intended context. Physical venues, partners and installations need separate confirmation; no external venue partnership is claimed.
- Credit every contributor with permission. Publish measured results and limitations only after human review; preserve a record of corrections.
Inspired by the interdisciplinary spirit of Bell Labs. This is an original, independent programme with no affiliation with or endorsement by Nokia Bell Labs, AT&T or their affiliates.
04 / Proposed seasons
2026 / First proofs
2026 proposal: choose one or two low-risk pilots, finalize scope and prepare safe fixtures.
2027 / Repeat + connect
2027 proposal: revisit successful pilots with mentors; publish only verified results and limitations.
No intern work or completed experiment is represented as published here. This shelf can hold credited evidence after contributor consent and human review.