Aprendizados sobre IA e Educação

Daily briefing · 7/6/2026 · 4 entities
Why I'm seeing this: An open request is looking for a specialist for an AI in education course.
connectionopen — you can help
Request for AI education specialist
expertise
Looking for: a teacher or specialist for an educational management course with AI

Action: Reply if you can suggest a teacher or specialist.

"me pediram indicação de um docente/especialista para apoiar num curso sobre Gestão Educacional com IA. Indicações? Candidatos?"

Natalia, Tanci Simões
Why I'm seeing this: A live education webinar is scheduled for next Monday.
event
ANPEd live on AI and teaching
📅 next Monday at 19:00📍 ANPEd YouTube channelonlinewebinar

Action: Watch on the ANPEd YouTube channel next Monday at 19:00.

"Formar humanos em tempos de aprendizagem com máquinas: desafios contemporâneos da docência e da mediação"

Sissi Lucena
Why I'm seeing this: A practical AI-creation story may be worth saving as an example of bottom-up adoption.
resource
Exame article on AI app builder
articleartificial intelligence
An Exame article about a mother who built an AI app and reached 50,000 accesses.

Action: Read the Exame article if you want a reference on grassroots AI apps.

"Sem formação em tecnologia, mãe desenvolve aplicativo com IA e atinge 50 mil acessos"

Luiz Eduardo Leao
Why I'm seeing this: A major education event got a new speaker update.
announcement
Educa Week 2026 speaker update

Action: Save Educa Week 2026 if you follow Brazilian education conferences.

"Compartilho uma nova atualização da Educa Week 2026, nossa edição especial de 10 anos."

Marcos Talarico MT, Eduardo Saron, Estela Zanini
4 entities not used
eventANPEd live on teaching and mediation in the age of machine learning
requestRequest for a specialist to support an educational management course with AI
resourceArticle on a mother building an AI app without formal tech training
announcementEduca Week 2026 gets a new speaker update
generated by zaplens · gpt-5.4-mini
debug — preferences, prompt and cost (remove later)
7,144 input tokens · 697 output tokens
Cost: US$ 0.0085 with cache · US$ 0.0085 without cache
entidades usadas (4 — todas de ontem)
event/webinar — ANPEd live on teaching and mediation in the age of machine learning
request/recommendation — Request for a specialist to support an educational management course with AI
resource/article — Article on a mother building an AI app without formal tech training
announcement/round — Educa Week 2026 gets a new speaker update
Preferences used (the lens)
No custom preferences set for this group — surface anything a member would genuinely not want to miss, and nothing else.

Include:
- Requests/asks someone made — especially ones still open (a member might help), and ones that just got answered.
- Offers, opportunities and useful resources shared: jobs, intros, tools, links, docs, discounts.
- Concrete decisions, announcements, plans, deadlines, dates and events.
- Facts or updates that actually change something for the group.

Ignore pure chatter with no concrete takeaway: greetings, reactions, small talk, memes, off-topic banter.
Prompt used
You write a reader's PERSONAL daily briefing of ONE WhatsApp group: only what matters to THEM, nothing else.
The input is NOT raw messages — it's ENTITIES already extracted from yesterday (structured episodes: a request/offer with state, or a dated fact). Your job is to SELECT the ones that match the reader's preferences and compose each into a briefing item.

LANGUAGE: the entities below may be in another language, but you MUST write EVERY human-readable output value (headline, summary, whyRelevant, action, and every type-specific field) in natural English, TRANSLATING the content. Do NOT echo the source language. Keep verbatim (do not translate): people's names, company/product names, links, the quote field, and dates/times.

GOLDEN RULE — be selective and honest:
- Include an entity ONLY if it matches the reader's preferences below DIRECTLY. The preferences are the ONLY inclusion criterion.
- The match must be PLAIN and PRESENT in the entity. Do NOT force, stretch, or invent relevance. If your justification needs ANY hedge — 'could', 'might', 'is an opportunity to', 'only if', 'unless', 'assuming', 'not directly' — then it does NOT match: DROP it.
- A day with nothing that matches is COMMON and EXPECTED. Returning items: [] is a SUCCESS, not a failure: the reader re-runs whenever they want, so a clean empty briefing costs nothing, while ONE off-topic item burns their trust. NEVER add an item just to fill the page.
- Do NOT broaden or re-interpret the reader's interest to fit an entity. A genuinely useful entity is still OFF-TOPIC if the lens didn't ask for that kind of thing: drop it.
- If the preferences explicitly list a category to IGNORE, exclude it even if it seems mildly useful. The preferences win.
- One entity ≈ one item, but MERGE entities that are the same story (e.g. a request and the entity that resolved it).

KEEP TEXT LEAN — favor short bullet phrases over paragraphs.

For each INCLUDED item, fill every field:
- headline: short and concrete (you may reuse/adapt the entity's headline).
- summary: 2-4 SHORT bullet phrases of what happened, naming the specific people/companies/links. From the entity's summary/detail.
- whyRelevant: a PLAIN, present-tense tie to a SPECIFIC preference, in the OUTPUT language. No hedging — if it needs 'if'/'might'/'unless', drop the item. NEVER first person.
- type — choose exactly ONE category; then fill THAT type's extra fields (the schema only accepts the matching ones):
    • intro — a new member who ACTUALLY presents themselves (a bare welcome/onboarding ping with no self-introduction is NOISE, not an intro — drop it). Extra: name, background, doing, lookingFor, location.
    • connection — someone asking for OR offering a contact/intro/expertise. Extra: direction (asking|offering), seeking (intro|contact|expertise|cofounder|hire|mentor|other), target (who/what the connection is to).
    • opportunity — a job, gig, deal, partnership or investment. Extra: opportunityKind, role, org, comp, location, contact.
    • recommendation — a specific vendor/professional/service/tool named to meet a need. Extra: need, category, recommended[] (each {name, what, vouchedBy}). If the ask is still unanswered, recommended is an empty array.
    • resource — a perk/link/survey/curated list/community/tool/article shared. Extra: resourceKind, what, topic.
    • event — a meetup/call/webinar/deadline/gathering. Extra: whenText (the date/time as said), startsAt & endsAt (ISO 8601, RESOLVED relative to the DATE of the message that mentioned it — null if you can't resolve it confidently), allDay, location, online, eventKind.
    • insight — a takeaway/decision/heuristic worth knowing. (no extra fields)
    • question — an open help request that is NOT about a contact. (no extra fields)
    • announcement — an org/community change. (no extra fields)
    • other — notable but fits none of the above. (no extra fields)
  ALL links go in the `links` field, for every type. For any type-specific field you cannot fill from the messages, use null — never guess.
- action: the concrete next step the READER can take, in the OUTPUT language. null if none.
- open: true ONLY when the entity is a request/ask still awaiting an answer the reader could provide (entity status 'open'). A resolved request, an announcement, or pure FYI = false.
- links: array of any URLs present in the entity; null if none.
- quote: the entity's verbatim quote if it adds value; null otherwise. Keep verbatim.
- senderNames: the entity's people. If someone is only a phone number, use the number — NEVER invent a name.
- relevance: 0..1. Order the items from most to least relevant.
- entityIds: the ids shown at the START of each entity line below (e.g. E1, E3) that you USED to compose this item. List every entity that fed the card. If you merge two entities into one card, list BOTH. Copy the ids verbatim.